# Flowful > AI-driven workflows and automation services by Flowful. This file contains public, indexable Flowful page content for LLM retrieval. Each page includes a source URL, language, and page description before the extracted content. ## English Pages ### About Flowful AI Source: https://flowful.ai/about/ Language: en Description: Meet the Flowful AI team: senior engineers (ex-Yelp, ex-SocGen), an AI developer and a sales lead. Custom AI for businesses in Belgium, France and Switzerland. AI CONSULTING & DEVELOPMENT > Flowful AI helps businesses automate work, improve decisions and deliver better customer experiences with production-ready AI. **Chatbots and AI assistants**Answer questions, route requests and handle support 24/7. **Process and workflow automation**Turn repetitive tasks into reliable workflows, integrated with your existing tools. **Custom AI applications and SaaS**Tailored to your business, from internal tools to customer-facing products. Located near [Brussels](/agency/brussels), [Lyon](/agency/lyon) and [Lausanne](/agency/lausanne), we work with clients across Europe and North America. Who we are #### The Team A partner committed to your success. We listen, understand and deliver beyond expectations. ##### Nicolas Chourrout Founder & AI Engineer Former Director of Engineering at Yelp with 17+ years building applications at scale. Focuses on architecture and turning AI ideas into production-ready systems. ##### Arnaud Jézéquel Co-Founder & AI Engineer 17+ years in market finance and IT at Société Générale CIB. Led major transformation programs including a multi-year platform overhaul. Deep expertise in data-intensive systems and regulated environments. ##### Manuel Temmerman Head of Sales 15+ years in business development, commercial strategy, and operations across Europe and Africa. Former General Manager at NUVIA Belgium (VINCI Group). Drives Flowful's growth and helps clients identify the right AI solutions. ##### Robin Fettelaar AI Developer Machine learning developer who built multimodal RAG systems and document automation at Dutch grid operator Alliander. BSc in Artificial Intelligence from Radboud University, where he was also a teaching assistant for deep learning and reinforcement learning. We're growing through partnerships and flexible collaborations. TRUSTED BY INNOVATIVE COMPANIES #### How We Work Clear scope, short cycles, real progress. We start small, prove value fast, and scale only what works. We build long-term partnerships, not one-off projects. Our work doesn't stop at launch. ##### Your goals drive the build We start from what matters to your business. Fewer tickets, faster processing, better conversion. The metric shapes the solution. ##### Prototype first, scale later A small working version on real data before committing to a full build. If it doesn't prove useful, we cut it early. ##### Built for production Security, monitoring, and resilience from day one. Every system we deliver runs reliably without constant attention. ##### Full visibility You always know what is automated, which data flows where, and how the system behaves. Your team stays in control. #### Ready to transform your business? Let's discuss how AI can help you achieve your goals. [Get in Touch](/contact) ### AI Agency Brussels Source: https://flowful.ai/agency/brussels/ Language: en Description: AI consulting and development near Brussels. Chatbots, workflow automation and custom AI apps for businesses in Belgium. Free 30-min consultation. AI CONSULTING & DEVELOPMENT AI consulting and development agency near Brussels. We build chatbots, workflow automations, and custom AI applications for businesses in Belgium. Free 30-min consultation. Our **AI development agency is located in Waterloo, in Walloon Brabant, at the gates of Brussels**. We support Belgian businesses in their digital transformation with concrete, measurable AI solutions. Based near Brussels, with teams in Lyon and Lausanne, our team combines technical expertise and business understanding to create innovative and efficient solutions. #### AI automation for businesses in Belgium We develop **artificial intelligence solutions for Belgian businesses** across various sectors: finance, healthcare, e-commerce, industry, and services. ##### Business process automation Custom workflows tailored to your tools ##### Intelligent chatbots Improve customer experience and assist your teams ##### Recommendation systems Advanced data analysis and recommendation tools ##### Web applications & APIs Integrate AI into your existing tools ##### Email automation AI-powered email processing and auto-responses Each project is designed to maximize return on investment while ensuring data security and confidentiality. Find more information about [our projects here](/projects/). #### AI funding and subsidies in Belgium Several **public funding programs exist for AI projects in Belgium**: regional subsidies (Innoviris, Digital Wallonia, VLAIO), enterprise vouchers, and federal tax incentives. Amounts and conditions vary by region and project type. Learn more in our [complete guide to AI funding in Belgium](/blog/fund-ai-project-belgium/). ##### Regional subsidies Brussels, Wallonia, and Flanders each offer dedicated programs for AI and digital transformation ##### Federal tax incentives R\&D deductions, innovation income deductions, and withholding tax exemptions #### Technical expertise in machine learning and generative AI Our **development and AI engineering team** masters the latest technologies: LLM, RAG, multi-agent systems, workflow orchestration, API development, data processing pipelines, machine learning, and chatbot development. Thanks to this expertise, our Waterloo agency designs high-performance and innovative solutions tailored to the specific needs of each Belgian business. #### Comprehensive support and strategic consulting An optional **AI audit** pinpoints where automation will have the most impact. We handle the full cycle: architecture, development, deployment, and iteration. We also train your teams on the tools we deliver so they gain autonomy over time. #### Why choose our AI agency in Belgium Working with our **AI development agency in Waterloo** means benefiting from a close, reliable, and innovative partner. We understand the specific challenges of businesses in Wallonia, Brussels, and Flanders and adapt our solutions to your needs. With 28+ projects delivered and 970+ hours saved per month for our clients, artificial intelligence becomes a strategic lever to accelerate your growth, optimize your business processes, and strengthen your competitiveness in the Belgian and European market. #### Your contacts ##### Nicolas Chourrout Founder & AI Engineer Former Director of Engineering at Yelp with 17+ years building applications at scale. Focuses on architecture and turning AI ideas into production-ready systems. ##### Manuel Temmerman Head of Sales 15+ years in business development, commercial strategy, and operations across Europe and Africa. Former General Manager at NUVIA Belgium (VINCI Group). Drives Flowful's growth and helps clients identify the right AI solutions. #### Frequently asked questions about our Brussels agency Where exactly are you based, and can you work on site in Brussels? Our office is in Waterloo, in Brabant wallon, about 20 km south of Brussels. We are available for meetings in Brabant wallon and the Brussels region, and can travel further when a project calls for it. The rest of the work happens remotely, so distance never sets the pace of a project. Which languages do you work in? English and French across the team, and Dutch with several of us. We localise most of what we ship into all three: a support chatbot or an email assistant that only handles one language solves half the problem in this market. Our own packages, documentation and this site exist in all three, and the assistants we build are multilingual by default rather than as a paid extra. How do we start, and what does a first project cost? Three entry points, depending on how much is already decided. An optional [AI Audit](/audit/) maps your processes and scores them by return, and half of its cost is credited toward whatever you build next. A [Proof of Concept](/proof-of-concept/) runs one to two weeks on your real data and answers a single question: does this actually work here? Or you take a [ready-made package](/packages/) from €1,000 setup. Every route starts with a free 30-minute call. Where does our data live, and how do you handle GDPR? Your data is encrypted, never used to train AI models, and stays under your control. A Data Processing Agreement comes as standard. Some components currently run on US infrastructure; EU-only hosting is available on request. On-premise deployment is also possible. If you are wondering what the AI Act asks of you specifically, we wrote [a practical guide for Belgian SMEs](/blog/eu-ai-act-compliance-smes/). Can Belgian subsidies cover part of an AI project? Often, though it depends on your region, and the programmes and their open calls change from one year to the next. Brussels, Wallonia and Flanders each run their own, and federal R\&D deductions apply on top of regional support. Rather than list conditions that date quickly, we keep the current picture, and what actually gets approved, in our [guide to funding an AI project in Belgium](/blog/fund-ai-project-belgium/). Which Belgian companies have you already worked with? Belgian companies run through a good part of our portfolio. Among them: [Oskar Architecten](/projects/oskar-internal-chatbot/), an internal knowledge chatbot and custom agents for an architecture practice; [La Maison Belge du Crédit](/projects/mbc-customer-automation/), a customer service automation suite; [Ixina](/projects/ixina-visual-quote-enricher/), enriching kitchen quotes with product images; [A2Com](/projects/a2com-ai-web-audit-tool/), an AI web audit tool; and [OldFashiond](/projects/oldfashiond-ai-matching-backend/), an AI-powered matching backend. Several are written up in full among [our projects](/projects/). #### Ready to discuss your project? Book a free 30-minute call with us to explore how AI can help your business. [Contact Us](/contact) ### AI Agency Lausanne Source: https://flowful.ai/agency/lausanne/ Language: en Description: AI consulting and development for Swiss companies, from Lausanne. Chatbots, workflow automation and custom AI, hosted in Switzerland or on your hardware. AI CONSULTING & DEVELOPMENT AI consulting and development for Swiss companies, from Lausanne. Chatbots, workflow automation and custom AI, hosted in Switzerland or on your own hardware. Free 30-min consultation. Flowful is **present in Lausanne**, on the shores of Lake Geneva, alongside our teams near Brussels and in Lyon. We work with Swiss companies that want artificial intelligence they actually control: automations, assistants and custom applications, built on infrastructure you choose rather than whichever cloud the vendor happens to resell. #### AI automation for companies in Switzerland We build **artificial intelligence solutions for Swiss companies** across the Arc lémanique and beyond: finance, medtech, industry, hospitality, and professional services. ##### Business process automation Custom workflows tailored to the tools you already run ##### Intelligent chatbots Customer support and internal knowledge, multilingual by default ##### Data analysis and recommendation Turn the data you already hold into decisions ##### Web applications & APIs Integrate AI into your existing tools ##### Email automation AI-powered email processing and auto-responses Every project is designed for measurable return, and for a hosting model you sign off on before a line of code is written. Find more information about [our projects here](/projects/). #### Sovereign AI, from Swiss cloud to your own server room Our **AI engineering team** works with LLM, RAG, multi-agent systems, workflow orchestration, API development, data pipelines and chatbot design. What changes on Swiss projects is **where the model runs**: open-weight models such as Mistral and Llama can be served from a Swiss datacenter, or from hardware inside your own building, so no prompt, document or customer record crosses a border you did not pick. #### Comprehensive support and strategic consulting An optional **AI audit** pinpoints where automation will have the most impact. We handle the full cycle: architecture, development, deployment, and iteration. We also train your teams on the tools we deliver so they gain autonomy over time. #### Why work with our AI team in Switzerland Working with **our AI team in Lausanne** means a partner who is close enough to meet in Lausanne and small enough to answer the phone. We understand what Swiss companies are weighing up, because they raise it in the first meeting: not whether AI works, but who ends up holding the data. With 28+ projects delivered and 970+ hours saved per month for our clients, artificial intelligence becomes a lever for growth and efficiency, on a deployment that answers to Swiss law rather than to a hyperscaler's terms of service. #### Your contact in Lausanne ##### Manuel Temmerman Head of Sales 15+ years in business development, commercial strategy, and operations across Europe and Africa. Former General Manager at NUVIA Belgium (VINCI Group). Drives Flowful's growth and helps clients identify the right AI solutions. #### Frequently asked questions about our Lausanne presence Are you actually in Switzerland, or is this run from abroad? Part of our team is based in Lausanne. Day to day the work happens over video, exactly as it does for our clients [near Brussels](/agency/brussels/) and [in Lyon](/agency/lyon/), so distance never sets the pace of a project. When meeting in person is worth it, we do that in Lausanne; travel further afield is arranged case by case and budgeted into the project. Where does our data live, and does any of it have to leave Switzerland? That is your decision, and we make it with you before we build. Three routes. **Swiss hosting**: providers such as Exoscale, itself headquartered in Lausanne, and Infomaniak run Swiss-only infrastructure under Swiss law, with no US parent company in the chain. **EU-only hosting**, if your customers sit mostly across the border. Or **fully on-premise**, where open-weight models run on hardware in your own server room and nothing leaves the building. In all three, data is encrypted, never used to train models, and a Data Processing Agreement comes as standard. Is a self-hosted model good enough, or are we giving up quality for control? You give up some, and it is a smaller gap than it was. Open-weight models now handle document search, drafting, classification and internal Q\&A at a level that holds up in daily use; the frontier hosted models still lead on the hardest reasoning. So we split it: what touches sensitive data runs on your infrastructure, and anything that does not can use a hosted model if the quality is worth it. Our note on [centralising your company's AI](/blog/from-chaos-to-control-centralizing-your-companys-ai/) covers why that boundary is far easier to hold with one system than with nine tools each doing their own thing. Does the EU AI Act apply to a Swiss company? Not directly, but it reaches past the border: if you place an AI system on the EU market, or its output is used there, you are in scope. Switzerland took a different route, signing the Council of Europe's AI convention in March 2025 and opting for sector-specific rules rather than one act, with a consultation draft due by the end of 2026. The revised Federal Act on Data Protection has applied since September 2023 regardless. Our [practical guide to the AI Act](/blog/eu-ai-act-compliance-smes/) covers the European dates that actually bind, which is what matters for anything you sell across the border. How do we start, and what does a first project cost? Three entry points, depending on how much is already decided. An optional [AI Audit](/audit/) maps your processes and scores them by return, and half of its cost is credited toward whatever you build next. A [Proof of Concept](/proof-of-concept/) runs one to two weeks on your real data and answers a single question: does this actually work here? Or you take a [ready-made package](/packages/) from €1,000 setup. Every route starts with a free 30-minute call. Switzerland is new for you. What have you actually built? Switzerland is our newest market and we would rather say so than pretend otherwise. The engineering behind it is not new. Among others: [Oskar Architecten](/projects/oskar-internal-chatbot/), an internal knowledge chatbot and custom agents for an architecture practice; [La Maison Belge du Crédit](/projects/mbc-customer-automation/), a customer service automation suite in a regulated sector; and [Elée](/projects/elee-multi-agent-powerpoint-presentation-generator/), a multi-agent generator that produces client-ready decks. Our [full project list](/projects/) spans Belgium, France and beyond. #### Ready to discuss your project? Book a free 30-minute call with us to explore how AI can help your business. [Contact Us](/contact) ### AI Agency Lyon Source: https://flowful.ai/agency/lyon/ Language: en Description: AI consulting and development agency in Lyon. Chatbots, workflow automation and custom AI apps for businesses in Auvergne-Rhône-Alpes and across France. AI CONSULTING & DEVELOPMENT AI consulting and development agency in Lyon. We build chatbots, workflow automations, and custom AI applications for businesses in Auvergne-Rhône-Alpes and across France. Free 30-min consultation. Our **AI development agency in Lyon** supports businesses in their digital evolution. We design and deploy tailor-made AI solutions to automate processes, improve operational performance, and enhance customer experience. Founded by two Lyon natives, with an office near Brussels (in Waterloo, Walloon Brabant) and a presence in Lausanne, our team combines technological expertise and business vision to deliver concrete, high-performing, and sustainable projects. #### AI automation for businesses in France We design **artificial intelligence applications** for many business areas, including finance, healthcare, e-commerce, industry, and services. ##### Intelligent automation Internal processes with custom workflows tailored to your tools ##### Advanced chatbots Optimize customer relations and support teams ##### Recommendation engines High-value data analysis and recommendation tools ##### Web applications & APIs Integrate AI into your existing tools ##### Email automation AI-powered email processing and auto-responses Each solution is designed to generate measurable return on investment while meeting security and data protection requirements. Discover our achievements on our [projects page](/projects/). #### AI project funding in France France offers several **schemes to fund your AI projects**. We can help you navigate these funding options and connect you with the right resources. Learn more in our [complete guide to AI funding in France](/blog/fund-ai-project-france/). ##### Crédit d'Impôt Recherche (CIR) Tax deduction on qualifying R\&D expenses, applicable to AI projects ##### France 2030 National investment plan with dedicated AI and digital tracks ##### BPI France Loans and grants for innovative SME and mid-cap projects ##### Regional grants Auvergne-Rhône-Alpes support for innovation and digital transformation #### Advanced expertise in machine learning and generative AI Our **AI engineering team** uses the latest market technologies: LLM, RAG, multi-agent architectures, automated workflow orchestration, API development, data processing pipelines, machine learning, and intelligent chatbot design. Thanks to this expertise, our Lyon-based agency designs robust, scalable solutions perfectly integrated into your existing IT environment. #### Strategic consulting and end-to-end support An optional **AI audit** identifies where automation delivers real value. From there, we handle architecture, development, deployment, and ongoing iteration. We train your teams on the tools we build so they can manage and extend them without depending on us. #### Why trust our AI agency in Lyon Choosing **our AI development agency in Lyon** means surrounding yourself with a committed, innovative, and attentive partner. We understand the challenges of businesses in Auvergne-Rhône-Alpes and beyond, and adapt each solution to your specific objectives. With 28+ projects delivered and 970+ hours saved per month for our clients, AI becomes a real lever for growth, business optimization, and competitive differentiation in the French and European market. #### Your contact in Lyon ##### Arnaud Jézéquel Co-Founder & AI Engineer 17+ years in market finance and IT at Société Générale CIB. Led major transformation programs including a multi-year platform overhaul. Deep expertise in data-intensive systems and regulated environments. #### Frequently asked questions about our Lyon agency Are you actually in Lyon, or is this a satellite address? Flowful was founded by two Lyon natives, and the French entity is based in Dardilly, in the north-west of Grand Lyon, where it is a registered member of [Techlid](https://www.techlid.fr/entreprises/flowful-ai/), the area's economic development association. We are available for meetings across Grand Lyon and the Rhône, and can travel further when a project calls for it. The rest of the work happens remotely, so distance never sets the pace of a project. Which languages do you work in? French and English day to day, plus Dutch through the Belgian side of the team. The assistants we build are multilingual by default rather than as a paid extra, which matters as soon as you serve customers outside France. How do we start, and what does a first project cost? Three entry points, depending on how much is already decided. An optional [AI Audit](/audit/) maps your processes and scores them by return, and half of its cost is credited toward whatever you build next. A [Proof of Concept](/proof-of-concept/) runs one to two weeks on your real data and answers a single question: does this actually work here? Or you take a [ready-made package](/packages/) from €1,000 setup. Every route starts with a free 30-minute call. Where does our data live, and how do you handle GDPR? Your data is encrypted, never used to train AI models, and stays under your control. A Data Processing Agreement comes as standard. Some components currently run on US infrastructure; EU-only hosting is available on request. On-premise deployment is also possible. The AI Act applies to French companies on the same timetable as Belgian ones, and our [practical guide](/blog/eu-ai-act-compliance-smes/) covers the dates that actually bind. Can French funding cover part of an AI project? Frequently. The Crédit d'Impôt Recherche covers qualifying R\&D spend, the Crédit d'Impôt Innovation covers prototyping for SMEs, BPI France funds innovative projects with loans and grants, France 2030 runs dedicated AI tracks, and Auvergne-Rhône-Alpes adds regional support for digital transformation. Which one fits depends on whether your project is research, prototyping or deployment, and the rates and eligibility rules move from one year to the next. Our [guide to funding an AI project in France](/blog/fund-ai-project-france/) walks through each, including what makes an application fail. Which French companies have you already worked with? Among others: [Elée](/projects/elee-multi-agent-powerpoint-presentation-generator/), a multi-agent generator that produces client-ready PowerPoint decks; [CTC](/projects/ctc/), AI resume analysis with a professional website; and [CT RH](/projects/ctrh/), a site for an HR consultancy. Our [full project list](/projects/) spans France, Belgium and beyond. #### Ready to discuss your project? Book a free 30-minute call with us to explore how AI can help your business. [Contact Us](/contact) ### AI Audit Source: https://flowful.ai/audit/ Language: en Description: A one-day remote AI audit: stakeholder interviews, process mapping, ROI scoring, a prioritized roadmap. €1,000, half of it credited to your next project. [All Packages](/packages) ASSESSMENT Not sure where to start? A one-day remote audit: half a day interviewing your team, half a day exploring your systems and drafting a written report with a prioritized roadmap. GDPR-ready EU-first hosting DPA available Why Audit First #### Know where AI pays off before you spend. Everyone is talking about AI. Very few know where it actually moves the needle in their business. One remote day of structured analysis replaces months of vague conversations and produces a clear, written verdict on what is worth building. - ✓Half a day on scheduled video calls with members of your team - ✓Half a day exploring your systems (via temporary read-only access, or documentation when direct access isn't possible) and drafting the written report - ✓Process map, opportunity scoring, ROI estimates, prioritized roadmap - ✓Recommendations go beyond AI: IT, internal processes, and website when it applies - ✓50% of the audit fee credited toward your next Flowful project - ✓Remote by default. On-site days in Belgium or France on request, with travel fees on top. Fixed Price ##### AI Audit €1,000flat - 1 day - Remote by default (on-site on request, travel fees extra) - Written report - 50% credited toward next project Flat price aimed at small organizations. Larger teams or more complex systems may need a longer audit, quoted separately during the scoping call. Deliverables #### What You Walk Away With Tangible output you can share with your team, board, or funding partner the day after the audit. ##### Stakeholder Interviews Structured conversations with operators, managers, and power users. We listen for friction, duplication, and tasks that drain time. ##### Process & Tooling Map A clear view of how work flows today: the tools, the handoffs, the workarounds. This is the baseline for every recommendation. ##### AI Opportunity Scoring with ROI Each opportunity is scored on impact, feasibility, and time-to-value, with a conservative ROI estimate. No hype, just numbers. ##### Prioritized Roadmap Quick wins you can ship in weeks, longer-term bets you can plan for, and the things to explicitly not do. A written document you own. ##### Recommendations Beyond AI We also flag what to fix in your IT, internal processes, and website when it applies. The goal is what actually moves the needle, not just the AI-shaped piece of it. How It Works #### From Kickoff to Report in a Week 1 ##### Scoping Call 30-minute free call to understand your business, confirm the audit is a fit, and agree on who we interview. 2 ##### Morning: Video Calls Half a day of scheduled video calls with members of your team, closest to the work. Tools, data, workflows, friction points. 3 ##### Afternoon: Exploration & Report Half a day exploring your systems (read-only access or documentation you share), then drafting the report with sized opportunities and recommendations. 4 ##### Debrief Once the report is finalized, we share it and walk through it together on a 1-hour debrief call. FAQ #### Frequently Asked Questions Who is the audit for? Teams with 5 to 200 people who suspect AI could help but want a clear, honest diagnosis before spending. Marketing, operations, support, sales, finance: any function with repetitive or knowledge-heavy work is a good candidate. Is the audit remote or on-site? Remote by default, run over scheduled video calls with members of your team. That is what lets us fit it in one day and keep the price flat. On-site days are available on request in Belgium, France and Switzerland, and quoted with travel fees (transport and, when required, accommodation) on top of the standard audit fee. Ask us before booking if you'd like an on-site quote. What does the 50% credit mean? If you decide to move forward with a Flowful project within 6 months of the audit, half of the audit fee is deducted from the first invoice of that project. Do the recommendations cover only AI? No. The audit focuses on AI opportunities, but when we spot issues in your IT, internal processes, or website that are blocking results (or cheaper to fix than an AI project), we call them out in the same report. Who do you talk to during the interviews? Typically a sponsor (founder or department lead), two or three operators who live in the workflow every day, and optionally a technical person if we need to understand existing systems. We scope this together before the audit. Do we need to prepare anything? Nothing heavy. We ask for a shared calendar slot for the morning video calls, and temporary read-only access to the systems we'll explore in the afternoon (CRM, helpdesk, knowledge base, analytics, whatever is relevant). If a direct read-only access isn't possible, we'll work from the documentation and exports you can share instead. We handle the rest. Can the audit cover compliance and EU AI Act readiness? Yes, on request. If you operate in the EU and want an AI Act risk review alongside the opportunity analysis, we can bundle it into the same day. #### Book Your AI Audit One day, written report, no fluff. 50% of the fee goes toward your next project. [Book an Audit](/contact) ### AI & automation blog Source: https://flowful.ai/blog/ Language: en Description: Insights on AI, automation, and intelligent solutions for business. Practical guides, case studies, and trends. Browse all articles. AI CONSULTING & DEVELOPMENT Insights on AI, automation, and building intelligent solutions #### [Sovereign AI in Europe: Mistral, EU Clouds and Your Options](/blog/sovereign-ai-mistral-eu-cloud) [Where your AI runs now matters as much as what it does. Four ways to keep company data under European control, from Mistral to EU clouds to your own server.](/blog/sovereign-ai-mistral-eu-cloud) [Read more →](/blog/sovereign-ai-mistral-eu-cloud) #### [Connect ChatGPT or Claude to Your Company Tools with MCP](/blog/connect-claude-chatgpt-internal-tools) [MCP connectors let AI assistants read your CRM, drives and tickets. What that unlocks for an SME, what it costs, and the three questions to settle first.](/blog/connect-claude-chatgpt-internal-tools) [Read more →](/blog/connect-claude-chatgpt-internal-tools) #### [AI Email Automation Explained: How a System Learns to Answer Email Like You](/blog/email-automation-ai-workflows) [Inside a production AI email system: extracting tone and FAQs from past conversations, knowledge bases and tools, templates, threads, signatures, and Gmail, Outlook or SMTP.](/blog/email-automation-ai-workflows) [Read more →](/blog/email-automation-ai-workflows) #### [EU AI Act Compliance for Belgian SMEs: What You Actually Need to Do](/blog/eu-ai-act-compliance-smes) [A practical guide to the EU AI Act for Belgian SMEs, updated after 2 August 2026: what Article 50 now requires from chatbots and voice agents, what the Digital Omnibus delayed, and a compliance checklist.](/blog/eu-ai-act-compliance-smes) [Read more →](/blog/eu-ai-act-compliance-smes) #### [Inside Vectoria: Our Unified AI Platform](/blog/inside-vectoria-one-brain-for-every-ai-package) [Why we built a unified AI platform instead of reinventing the wheel on every project, and what Vectoria actually gives you.](/blog/inside-vectoria-one-brain-for-every-ai-package) [Read more →](/blog/inside-vectoria-one-brain-for-every-ai-package) #### [How to Fund Your AI Project in France: Grants, Subsidies, and Tax Credits](/blog/fund-ai-project-france) [A practical guide to French funding for AI projects. CIR, CII, BPI France, France 2030, regional grants, and how to get your application approved.](/blog/fund-ai-project-france) [Read more →](/blog/fund-ai-project-france) #### [AI Readiness Checklist: Is Your Business Ready to Automate?](/blog/ai-readiness-checklist) [A practical checklist to evaluate whether your business is ready for AI automation. Data, processes, team, and budget: what you actually need before starting.](/blog/ai-readiness-checklist) [Read more →](/blog/ai-readiness-checklist) #### [How to Fund Your AI Project in Belgium: Subsidies, Grants, and Tax Credits](/blog/fund-ai-project-belgium) [Belgian funding for AI and digital projects, checked September 2026: Prime Digitalisation, Chèques-Entreprises, Start IA, Tremplin IA, VLAIO, tax incentives.](/blog/fund-ai-project-belgium) [Read more →](/blog/fund-ai-project-belgium) #### [GEO & AEO: How to Get Cited by ChatGPT, Perplexity, and AI Search Engines](/blog/geo-get-cited-by-chatgpt-ai-engines) [A 10-point checklist to get your business cited by ChatGPT, Perplexity, and AI search engines, from robots.txt and llms.txt to FAQ pages and E-E-A-T signals.](/blog/geo-get-cited-by-chatgpt-ai-engines) [Read more →](/blog/geo-get-cited-by-chatgpt-ai-engines) #### [AI Glossary: 15 Terms You Need to Know in 2026](/blog/ai-glossary-2026) [No PhD required. A plain-language guide to the AI terms you'll actually encounter in meetings, demos, and vendor pitches.](/blog/ai-glossary-2026) [Read more →](/blog/ai-glossary-2026) #### [Custom Tools vs Enterprise SaaS: Why Building Beats Configuring](/blog/custom-tools-vs-enterprise-saas) [Building custom internal tools is now faster than configuring generic SaaS. Why senior developers embrace AI coding, and what it means for your business.](/blog/custom-tools-vs-enterprise-saas) [Read more →](/blog/custom-tools-vs-enterprise-saas) #### [Voice Agents in 2026: Why They're Finally Ready](/blog/voice-agents-2026) [Voice AI has reached a tipping point: sub-100ms latency, native audio reasoning, and seamless workflow integration. Here is what changed in 2026.](/blog/voice-agents-2026) [Read more →](/blog/voice-agents-2026) #### [From Chaos to Control: Centralizing Your Company’s AI](/blog/from-chaos-to-control-centralizing-your-companys-ai) [Replace scattered ChatGPT accounts with a secure, centralized internal chatbot.](/blog/from-chaos-to-control-centralizing-your-companys-ai) [Read more →](/blog/from-chaos-to-control-centralizing-your-companys-ai) #### [Proof of Concept: The Smart Way to Start Your AI Project](/blog/proof-of-concept-the-smart-way-to-start-your-ai-project) [When prototyping beats weeks of guessing](/blog/proof-of-concept-the-smart-way-to-start-your-ai-project) [Read more →](/blog/proof-of-concept-the-smart-way-to-start-your-ai-project) #### [Can AI Go Green?](/blog/can-ai-go-green) [AI’s Climate Dilemma: Problem or Solution?](/blog/can-ai-go-green) [Read more →](/blog/can-ai-go-green) #### [Discover How AI Chatbots Can Elevate Your Customer Experience](/blog/why-your-business-needs-a-chatbot) [Save time and enhance service with AI-powered chatbots](/blog/why-your-business-needs-a-chatbot) [Read more →](/blog/why-your-business-needs-a-chatbot) #### [AI You Can Trust: Achieving Consistent Results](/blog/reliable-ai) [Learn simple solutions to make AI more reliable for your business](/blog/reliable-ai) [Read more →](/blog/reliable-ai) ### AI Glossary: 15 Terms You Need to Know in 2026 Source: https://flowful.ai/blog/ai-glossary-2026/ Language: en Description: No PhD required. A plain-language guide to the AI terms you'll actually encounter in meetings, demos, and vendor pitches. You’ve sat through a demo where someone casually dropped “RAG pipeline,” nodded along when a vendor mentioned “knowledge base,” and quietly Googled “what is an LLM” under the table. No shame: the AI world moves fast, and the vocabulary moves faster. This post is your cheat sheet. Fifteen terms, zero jargon rabbit holes. I’ll explain each one the way I’d explain it to a friend over coffee. --- #### Part 1: The Building Blocks These two concepts come up in nearly every AI conversation. Nail them and the rest clicks into place much faster. ##### 1. LLM - The engine behind modern AI tools LLM stands for Large Language Model. It’s a type of AI trained on massive amounts of text to understand and generate human language. At its core, it does one thing: **predict the next token** (the small chunks of text we’ll cover in #2). Give it “all that glitters” and it predicts “is not gold.” That simple idea, scaled up to billions of examples, produces something that feels like understanding. You probably know the products built on top of LLMs: ChatGPT, Claude, Gemini. These are AI assistants, each backed by model families from their respective providers (OpenAI, Anthropic, Google, Meta, and others). Under the hood, each provider offers multiple models tuned for different tradeoffs between speed, cost, and capability. **A note on cost:** More capable models cost more per use. For simple tasks like summarizing emails or answering FAQs, smaller, cheaper models (like Claude Haiku or Gemini Flash) work great and can cost a fraction of a cent per request. Save the heavy hitters for complex work. When someone says “we’re building on top of an LLM,” they mean they’re using one of these models as the engine behind their product. ##### 2. Token - The unit AI reads in, and charges you for AI models don’t read full words. They break text into subword chunks called tokens. Each model has its own tokenizer, so the exact split varies, but here’s roughly how it works: > **“Artificial intelligence is transforming business”** → `[Artific] [ial] [ intelligence] [ is] [ transform] [ing] [ business]` In English, a token averages about three-quarters of a word, though this varies across languages and models. Tokens are an implementation detail, not a linguistic unit, but they matter for two practical reasons: AI providers **charge per token**, and every model has a maximum number of tokens it can process at once (more on that in a moment). --- #### Part 2: Working with AI These are the concepts you’ll encounter every time you interact with or deploy an AI system. ##### 3. Prompt Engineering - The art of asking AI the right question Every time you type something into ChatGPT, that’s a prompt. But a well-crafted prompt includes context, examples, and specific instructions. It’s the difference between telling a new hire “handle this” and giving them a detailed brief with examples of what good looks like. In practice, prompt engineering usually means building **reusable prompt templates** with structured instructions, constraints, and output formats, not just clever one-off phrasing. One powerful technique is **few-shot prompting**, where you give the AI examples before asking your question: ```text Classify these support tickets: "My order hasn't arrived" → Shipping "The app keeps crashing" → Technical Now classify: "I was charged twice" → ? ``` By seeing the pattern, the AI gives you much more consistent results. ##### 4. Context Window - How much the AI can see at once Every AI model has a fixed token budget called the **context window**. Your input, any background documents, and the model’s own response all share this budget. Think of it as the size of the AI’s desk: the bigger the desk, the more documents it can spread out and reference at the same time. Early models (around 2022-2023) topped out around 4,000-8,000 tokens, roughly a few pages of text. Some modern models handle 128,000 to over 1 million tokens (enough for hundreds of pages), though the largest windows often come with higher-tier plans or API pricing. But there’s always a limit. The model itself only sees what fits in the window; when a conversation grows too long, the application in front of it may summarize, trim, or selectively retrieve older information to stay within budget. **Why this matters for cost:** In a chat conversation, every message you send includes the entire conversation history. The AI doesn’t just read your last message; it re-reads everything from the start. So the cost per message grows as the conversation gets longer. A 50-message thread costs significantly more per reply than a fresh one. This is why many applications automatically summarize or truncate older messages behind the scenes. ##### 5. Fine-Tuning - Teaching AI a specific style or behavior LLMs start as generalists: they know a little about everything. Fine-tuning trains them further on your data so they adopt a specific **tone, format, or behavior**. It’s less about adding new facts (RAG is better for that, see #10) and more about shaping _how_ the model responds. For example, you might fine-tune a model so it always uses your brand voice, follows a specific output format, or handles domain-specific terminology consistently. Modern approaches like LoRA (Low-Rank Adaptation) make this faster and cheaper than retraining a full model, but it still requires careful data preparation and evaluation. **In practice:** For most business use cases, RAG is a simpler and more cost-effective starting point. Fine-tuning shines when you need consistent behavior that’s hard to achieve through prompting alone. ##### 6. Hallucination - When AI confidently makes things up LLMs generate text based on patterns, not facts. Sometimes they produce something that sounds perfectly reasonable but is completely wrong: a fake statistic, a non-existent legal case, a citation that doesn’t exist. Think of that one colleague who always has an answer, even when they have no idea. Same energy. This is arguably **the most important term on this list** for anyone deploying AI. Hallucinations are an inherent risk of how these models work. They can be significantly reduced through grounding, retrieval (RAG), and verification layers, but not fully eliminated. If you’re putting AI in front of customers, you need safeguards in place. We wrote a deeper dive on this topic: **[How to Make AI Outputs Reliable](/blog/reliable-ai)**. ##### 7. Guardrails - Safety fences for your AI Guardrails are the rules, filters, and systems that keep your AI from going off the rails. In practice, this includes several layers: - **PII redaction**: automatically stripping personal data (names, emails, phone numbers) from inputs and outputs - **Policy filters**: blocking responses on off-limits topics or enforcing compliance rules - **Citation requirements**: forcing the AI to reference source documents rather than generating from memory - **Tool allow-lists**: restricting which actions an AI agent can take (e.g., read-only database access) - **Human-in-the-loop**: requiring manual approval before the AI takes high-stakes actions These range from simple prompt instructions to sophisticated multi-layer systems. If you’re putting AI in front of customers, guardrails aren’t optional. --- #### Part 3: Connecting AI to Your Data This is where things get exciting. These are the patterns that turn a generic AI into something that actually knows your business. ##### 8. Knowledge Base - Your company’s curated source of truth A knowledge base is the organized collection of documents, FAQs, policies, and data that you maintain as your AI’s reference material. Think of it as the employee handbook, product docs, and tribal knowledge all rolled into one curated library. The key word is _curated_. A knowledge base isn’t the same as the vector store or search index used to retrieve from it (that’s #9 and #10). It’s the source of truth you maintain: choosing what goes in, keeping it up to date, and structuring it so AI can find the right answer. Without a knowledge base, AI can only draw on its general training. With one, it becomes an expert on your company. When someone says “we need to build a knowledge base for the chatbot,” they mean collecting, organizing, and maintaining your internal information so the AI can reference it. ##### 9. Embeddings & Vector Databases - Organizing knowledge by meaning, not keywords When you add documents to a knowledge base, a specialized **embedding model** converts each chunk of text into an **embedding**: a list of numbers that captures its meaning. Imagine a map where every piece of content has a location. Similar ideas are close together, unrelated ones are far apart. A **vector database** is where these embeddings live. It’s like a library where books are organized by meaning, not title. When a customer asks a question, the vector database finds the most relevant content by proximity on that map. **Traditional database:** “Find me the row where `category = 'returns'`” (exact match) **Vector database:** “Find me everything _related to_ customer complaints about shipping” (meaning match) In practice, most systems combine both approaches: vector search for finding relevant content by meaning, plus traditional filters (by date, category, or source) to narrow results. One doesn’t replace the other. ##### 10. RAG - Giving AI a cheat sheet before it answers RAG stands for Retrieval-Augmented Generation. Before the AI answers your question, it first **searches your knowledge base** to find relevant information. Then it uses that information to generate a grounded answer. It’s the difference between an intern who’s winging it and one who checks the handbook first. RAG is how most businesses connect AI to their own data, and it’s usually a simpler starting point than fine-tuning (#5). **One important caveat:** RAG is only as good as its retrieval. If documents are poorly chunked, missing metadata, or the search returns the wrong passages, the AI will confidently answer from the wrong source. Getting retrieval right (chunking strategy, metadata, re-ranking) is often the hardest part of building a RAG system. ##### 11. MCP - A standard way to connect AI to your tools MCP stands for Model Context Protocol. It’s an open standard that defines how AI models connect to external tools and data sources. Before MCP, every integration had to be custom-built. If you wanted your CRM data available to both a customer chatbot and an internal email automation, the connecting logic had to be written twice, one for each AI interface. Think of it like USB-C for AI: a common interface that tools can implement so any compatible AI assistant can use them. Build one MCP server for your business data, and any MCP-compatible tool (your chatbot, your coding assistant, your email agent) can connect to it without duplicating work. The standard is still maturing, and adoption depends on both the AI tool and the service implementing it. Key areas like authentication and permission scopes are actively evolving. But the direction is clear: standardized integrations instead of one-off custom code. --- #### Part 4: The Frontier These concepts are shaping how AI is built and used today. ##### 12. Context Engineering - Designing what AI sees and when Context engineering is a relatively new term that describes the practice of designing the full information pipeline around an AI system. It goes beyond prompt engineering (#3) to encompass: - **Routing**: deciding which model or tool handles each request - **Memory management**: summarizing or trimming conversation history to stay within the token budget - **Retrieval orchestration**: pulling the right documents via RAG (#10) at the right moment - **Tool selection**: choosing which integrations (#11) to invoke - **Output formatting**: structuring responses for downstream systems If prompt engineering is writing a good email, context engineering is designing the entire communication workflow. The term isn’t universally standardized yet, but the practice is quickly becoming essential as AI systems grow more complex. ##### 13. AI Agent - AI that doesn’t just answer, it acts A regular AI chatbot answers questions. An AI agent can actually **do things**: it breaks down a task into steps, decides which tools to use, and executes. You may already be using one: Claude Code writes and runs code across entire projects, and ChatGPT’s “operator” mode can browse the web and fill out forms on your behalf. For a small business, an agent could [monitor your “info@” email address](/packages/email-automation), draft replies to common questions, and only alert you for the complex ones. Or it could process incoming invoices, match them against purchase orders, and flag discrepancies for review. The key difference from a chatbot: it takes action on your behalf, not just provides information. ##### 14. Multimodal - AI that can see, hear, and read Most early LLMs could only process text. Multimodal models can work with **multiple types of input and output**: text, images, audio, and video. Show it a photo of a damaged product and ask it to write a claim. Give it a chart and ask it to explain the trends. This is a bigger deal than it sounds. Before multimodal, building a [voice-based AI assistant](/packages/ai-phone-receptionist) meant chaining separate tools together: a speech-to-text service to transcribe, an LLM to generate a reply, and a text-to-speech service to read it back. Each step added latency and cost. A multimodal model handles the entire flow natively, which makes real-time voice conversations feel natural instead of stilted. ##### 15. Reasoning Model - AI that thinks before it answers Most AI assistants answer directly. Reasoning models spend extra compute working through a problem step by step before responding, checking assumptions and catching errors along the way. Some expose this thinking visibly (like DeepSeek R1), while others (like OpenAI’s o3) reason internally and return only the final answer. The tradeoff is time and cost for accuracy. Reasoning models shine for financial analysis, legal review, and data interpretation. For everyday tasks, standard models are faster and cheaper. --- #### Wrapping Up AI vocabulary can feel like a barrier, but most of these concepts map to ideas you already understand. The terms exist because the technology is genuinely new, not because anyone’s trying to confuse you. The next time a vendor pitches you an “agentic RAG solution with MCP integrations and multimodal reasoning,” you’ll know exactly what they mean, and more importantly, you’ll know the right questions to ask. Got a term I missed? Or one that still doesn’t make sense? **[Reach out to the Flowful team](/contact)**, we love helping businesses make sense of AI. ### AI Readiness Checklist: Is Your Business Ready to Automate? Source: https://flowful.ai/blog/ai-readiness-checklist/ Language: en Description: A practical checklist to evaluate whether your business is ready for AI automation. Data, processes, team, and budget: what you actually need before starting. Most AI projects don’t fail because the technology wasn’t good enough. They fail because the business wasn’t ready. Rate yourself in the scorecard below, then read the sections that matter most to you. #### Your Readiness Scorecard ##### How to read your score: - **Mostly “Ready”**: A PoC is the natural next step. - **Mix of “Ready” and “Getting There”**: Very common. You can start on the areas that are ready while improving the others. - **Mostly “Not Ready Yet”**: Now you know exactly what to work on first. Some of these gaps (like digitizing data or documenting processes) can be addressed in weeks, not months. Three areas are non-negotiable before starting: your data needs to be digital and accessible, you need someone internally who will own the project day-to-day, and you need a specific use case tied to a real problem. Everything else can be worked on as you go. Now, dig into each area below for the details. #### 1. Your Data AI runs on data. Not mountains of it, necessarily, but data that’s accessible, digital, and reasonably organized. Here’s what to evaluate. ##### Is your data digital? If your key records live in filing cabinets, handwritten notebooks, or scanned PDFs without OCR, you’ll need a digitization step before AI can touch them. That’s not a dealbreaker, but it adds time and cost. **If your data already lives in spreadsheets, a CRM, an ERP, or any database, you’re in much better shape.** ##### Do you have enough volume? You don’t need millions of records. For most business automation use cases (email classification, document extraction, customer routing), a few hundred representative examples is a solid starting point. With modern LLM-based approaches, you often need even less, since the model already understands language and context out of the box. If you’re looking at something more specialized like demand forecasting or anomaly detection, you’ll want several months of historical data. **The key question: do you have enough examples for someone (or something) to spot the pattern?** ##### Is it somewhat structured? A folder of random files named “final\_v3\_REAL.xlsx” is technically digital, but it’s a headache. If your data has consistent columns, labels, or categories, that’s a big plus. If it doesn’t, we can usually wrangle it into shape, but again, that’s extra work upfront. ##### Can you access it programmatically? Can we connect to your data source through an API, a database query, or an export? **If the only way to get the data is to manually copy-paste from a screen, that’s a bottleneck.** Most modern tools (Google Workspace, Microsoft 365, CRMs like HubSpot or Pipedrive, ERPs) offer APIs or integrations. If yours does, you’re in good shape. ##### Your data score: - All digital and in a database or CRM with API access? **Ready.** - Mostly in spreadsheets with some consistency? **Getting There.** Minor cleanup needed. - Mix of paper and digital, inconsistent formats? **Not Ready Yet.** Plan for a data preparation phase first. #### 2. Your Processes AI automates tasks, not magic. It works best when there’s a clear, repeatable process to build on. If nobody in the company can explain how something gets done today, an AI system won’t figure it out on its own. ##### Can you identify repetitive, rule-based tasks? The best candidates for automation are tasks someone does the same way, dozens or hundreds of times. Sorting incoming emails. Extracting data from invoices. Answering the same ten customer questions. Generating weekly reports from the same data sources. **If you can describe the task as “when X happens, do Y,” that’s a strong signal.** ##### Do you have documented processes (or at least consistent ones)? You don’t need a 50-page operations manual. But someone on your team should be able to walk through the steps of the process you want to automate. **If three people do the same task three different ways and nobody agrees on which way is “right,” that’s a process problem to solve before adding AI to it.** ##### Is there a clear human-in-the-loop step? The most successful AI automations keep a human in the loop, at least at the start. AI drafts the reply, a person reviews and sends it. AI extracts the invoice data, a person confirms before it hits the accounting system. Where does the human check happen in your process? **If you can define that clearly, the automation will be smoother and safer to deploy.** ##### What’s the cost of errors? If AI gets a customer response slightly wrong, you can fix it in the review step. If AI miscategorizes a high-priority support ticket and nobody catches it for three days, that’s a bigger problem. **Understand the stakes.** High-error-cost processes need tighter validation loops. Low-error-cost processes can move faster toward full automation. ##### Your process score: - Clear, repeatable tasks with documented steps and defined review points? **Ready.** - Consistent but undocumented processes? **Getting There.** We’ll document them together during setup. - Ad-hoc processes that vary wildly by person or situation? **Not Ready Yet.** Standardize the workflow before automating it. #### 3. Your Team Technology is the easy part. People are harder. An AI tool nobody uses is a waste of money, no matter how clever it is. ##### Do you have a champion or sponsor? The most successful AI projects we’ve delivered all had one thing in common: one person inside the company who pushed for it. Someone who understood the problem, believed in the approach, and had the authority (or the ear of someone with authority) to make decisions. This doesn’t need to be the CEO. It can be an operations manager, a team lead, or a department head. **But someone needs to own it.** ##### Is the team open to change? This is a real question, not a checkbox. If the people who will use the AI tool daily see it as a threat to their job rather than a tool that removes their least favorite tasks, adoption will be a struggle. **The best results come when teams are involved early**, understand what the tool will do (and won’t do), and have a say in how it works. ##### Who will own the AI tool day-to-day? After we build and deploy the automation, someone needs to monitor it, handle edge cases, and flag when something needs adjusting. **This doesn’t require a technical person.** It requires someone who understands the process and can spend 15 to 30 minutes a day reviewing outputs, especially in the first few weeks. Over time, that drops significantly as the system stabilizes. ##### Your team score: - Engaged champion, receptive team, clear day-to-day owner? **Ready.** - Interested leadership but the team hasn’t been involved yet? **Getting There.** Start with a demo or pilot to build buy-in. - Nobody’s really pushing for it, or the team is resistant? **Not Ready Yet.** Solve the people problem first. #### 4. Your Budget and Timeline AI projects don’t have to cost a fortune or take six months. But they do require an honest conversation about investment. ##### Start with a Proof of Concept. **A PoC takes 1 to 2 weeks and costs a fraction of a full build.** It answers the critical question: does this work for our data and our use case? From there, you decide whether to continue. We cover this in detail in [Proof of Concept: The Smart Way to Start Your AI Project](/blog/proof-of-concept-the-smart-way-to-start-your-ai-project). ##### Budget realistically. A focused AI automation ([email routing](/packages/email-automation), document extraction, a [customer-facing chatbot](/packages/web-chatbot)) typically runs in the low thousands for a PoC and low-to-mid five figures for a production build. **That’s a fraction of what most businesses spend annually on enterprise SaaS licenses they barely use.** For context, read [Custom Tools vs Enterprise SaaS: Why Building Beats Configuring](/blog/custom-tools-vs-enterprise-saas). Belgian companies can also tap into public funding. Innoviris (Brussels), Tremplin IA (Wallonia), or VLAIO (Flanders). See our [full guide on AI subsidies in Belgium](/blog/fund-ai-project-belgium). ##### Set realistic timeline expectations. From kickoff to a stable, running system, **most projects land in the 2 to 3 month range**: 1–2 weeks for a PoC, 4–8 weeks for production, and 2–4 weeks of stabilization. If someone promises you a fully autonomous AI system in two weeks, be skeptical. ##### Your budget score: - Budget allocated, timeline understood, open to starting with a PoC? **Ready.** - Budget exists but expectations need calibrating? **Getting There.** A scoping call will fix that quickly. - No budget yet, or expecting AI to be free and instant? **Not Ready Yet.** Have an honest conversation first. #### 5. Your Strategy This is the question most companies skip: do you actually know what you want AI to do? ##### Have you identified a specific use case? **“We want to use AI” is not a use case.** “We want to automatically classify incoming support emails by urgency and route them to the right team” is. The more specific you can be about the task, the input, and the expected output, the easier everything else becomes. If you’re not sure where to start, that’s fine, but recognizing that gap is step one. ##### Is the use case tied to a real business problem? The best AI projects solve a problem someone already feels. A team drowning in manual data entry. A support queue that takes too long. A reporting process that eats two days every month. **If you can point to a pain that costs time, money, or customer satisfaction, you’ve found your starting point.** If the use case is “because competitors are doing AI,” that’s not enough. ##### Can you measure success? Before starting, define what “working” looks like. How much time should the automation save? What accuracy is acceptable? What’s the current error rate you’re trying to beat? **Without a baseline and a target, you won’t know whether the project delivered value**. And neither will the person approving the budget. ##### Your strategy score: - Specific use case identified, tied to a measurable business problem? **Ready.** - General idea of where AI could help, but not yet specific? **Getting There.** A scoping session will sharpen that quickly. - No clear use case yet? **Not Ready Yet.** Start by listing your team’s most repetitive, time-consuming tasks. The use case is usually hiding in plain sight. #### 6. Your Governance You don’t need a 30-page AI policy to get started. But you do need to think about a few things before putting AI into production, especially in Europe. ##### Are you handling personal data? If the process you want to automate touches customer data, employee data, or any personally identifiable information, GDPR applies. That doesn’t mean you can’t use AI. It means you need to know what data flows where, ensure your AI provider has a proper Data Processing Agreement, and be clear about data retention. If you’re already GDPR-compliant in your current operations, extending that to an AI tool is usually straightforward. ##### Do you know where your data goes? When you use a cloud-based AI service, your data may be processed on external servers. Know which provider you’re using, where the servers are located (EU hosting matters), and whether your data is used to train their models. For sensitive business data, on-premise or private cloud options exist. **This isn’t about paranoia. It’s about making an informed choice.** ##### Have you thought about the EU AI Act? The [EU AI Act](https://artificialintelligenceact.eu/) entered into force in August 2024, with obligations being phased in through 2027. Prohibited AI practices have applied since February 2025, and rules for high-risk systems take effect from August 2026. Most business automation use cases (email routing, document processing, customer support) fall into the minimal or limited risk categories, which require only basic transparency measures. But if your use case involves decision-making that affects people (hiring, credit scoring, access to services), stricter rules apply. It’s worth a quick check before you build. ##### Your governance score: - GDPR-compliant, data flows understood, EU AI Act risk level checked? **Ready.** - GDPR basics in place but haven’t thought about AI-specific implications? **Getting There.** A brief review during project setup will handle it. - No data policies, unsure about compliance? **Not Ready Yet.** Address this before going live. It doesn’t take long, but it’s not optional. #### What’s Next? If you scored “Ready” or “Getting There” in at least three or four areas, you’re closer than you think. [Book a free scoping call](/contact) and we’ll walk through your specific situation, tell you what we’d actually build and in what order, and tell you honestly if now is the right time. No commitment, no prep needed. Still exploring? Read [Proof of Concept: The Smart Way to Start Your AI Project](/blog/proof-of-concept-the-smart-way-to-start-your-ai-project) or check out [Custom Tools vs Enterprise SaaS](/blog/custom-tools-vs-enterprise-saas). ### Can AI Go Green? Source: https://flowful.ai/blog/can-ai-go-green/ Language: en Description: AI’s Climate Dilemma: Problem or Solution? Many worry about the energy appetite of generative AI. [Sam Altman](https://blog.samaltman.com/the-gentle-singularity) recently wrote that a single ChatGPT query uses about **0.34 Wh**, roughly what a high‑efficiency LED burns in a couple of minutes. Scale that to billions of queries and the power draw of data centres quickly becomes a climate headline. A [Nature](https://www.nature.com/articles/s44168-025-00252-3) study published in June 2025 argues that smart, sector‑specific AI could cut **3.2–5.4 gigatonnes of CO₂‑equivalent per year (Gt CO2e) by 2035** across power, food and transport. A [Financial Times coverage](https://www.ft.com/content/bd835b8f-e39a-4e5f-84d0-2fb019b47b80) highlights the same upside but warns that real‑world deployment still lags the lab results. Below is a quick tour through the upside, the footprint, the hype filter and the imperatives for greener AI. #### 1. The upside: five levers worth about 5 gigatonnes - **Smarter grids and traffic**: AI predicts demand, balances renewables and unclogs city roads. - **Better batteries and less waste**: generative models speed up chemistry search and trim packaging. - **Greener everyday choices**: tools like Google Maps eco‑routes nudge users to save fuel. - **Sharper climate forecasts**: high‑resolution models warn of floods and droughts sooner. - **Faster disaster response:** early warnings for fires and storms protect people and assets. Focusing on power, food and mobility alone, Stern et al. estimate AI could avoid up to **5.4 Gt CO₂e** each year, roughly three times AI’s own projected emissions. Blue depicts business-as-usual with only modest cuts, green shows AI-enabled efficiency shaving 3–5 Gt CO₂ a year by 2035, and the dashed navy line represents ambitious climate policies delivering the steepest reductions (Source: Nature). #### 2. The footprint: rising fast The [IEA](https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centres-while-offering-the-potential-to-transform-how-the-energy-sector-works) predicts that data centre electricity demand will more than double to about **945 TWh by 2030**, with AI the biggest driver. That is similar to Japan’s entire grid today. Key points: - **Location matters:** Training on a Belgian winter night fed by offshore wind is cleaner than a sunny Texas afternoon on a coal-heavy grid. - **Water consumption is a hidden cost:** AI’s footprint isn’t just carbon. Data centres consume vast amounts of fresh water for cooling. The _Nature_ study highlights this, calling for transparent reporting of water usage alongside energy and emissions. - **Model size is not free:** A ten times larger model can need ten to twenty times more energy for marginal gains. - **Hardware and cooling:** Liquid cooling and faster GPUs buy time but do not solve the problem alone. #### 3. The hype filter: lessons from the FT The Financial Times’ Pilita Clark notes that prototypes wow investors, but messy field conditions and thin margins derail many climate AI pilots. Meta’s CO₂ capture model, for example, unravelled under peer review. #### 4. Green AI in practice: Four imperatives - **Track and disclose emissions** both sources call for transparent reports of model energy, carbon and water use. - **Aim at high impact sectors** deploy scarce GPUs where decarbonisation gains are largest, namely power, food and mobility. - **Test outside the lab** validate with live data and publish uncertainties to close the lab to market gap. - **Adopt efficient tech** NVIDIA Blackwell GPUs promise up to twenty five times better energy per token than H100s, and Mixture of Experts routing can cut inference energy by about seventy percent. #### 5. Policy: the active state angle Markets alone will not prioritise low profit high impact climate applications. The **Nature** paper suggests governments should: - Mandate lifecycle emissions disclosure for cloud and model providers. - Offer time of use rebates for AI workloads that align with renewable peaks. - Fund open climate AI datasets and benchmarks to de risk early research. #### Bottom line AI can be a climate multiplier, but only if we measure its footprint, steer GPUs toward the biggest carbon wins and adopt the efficiency tech already on the roadmap. Ready to stress test your own AI energy budget? [**Contact us**](/contact); we can measure your AI climate impact and optimise it for a lighter footprint. ### Connect ChatGPT or Claude to Your Company Tools with MCP Source: https://flowful.ai/blog/connect-claude-chatgpt-internal-tools/ Language: en Description: MCP connectors let AI assistants read your CRM, drives and tickets. What that unlocks for an SME, what it costs, and the three questions to settle first. Ask ChatGPT or Claude “what did we quote client X in March?” and you get a shrug: the answer sits in your CRM, drive or inbox, not in the model. In 2026, both assistants can connect to those tools. Here is what that unlocks, what it costs, and the three questions to settle first. #### What is MCP, and why it matters now The **[Model Context Protocol (MCP)](https://modelcontextprotocol.io)** is an open standard that lets an AI assistant read from, and act in, your other software through one common kind of connector. Anthropic [released it](https://www.anthropic.com/news/model-context-protocol) in late 2024, OpenAI adopted it in March 2025, and Google and Microsoft followed. In December 2025 Anthropic handed it to the [Agentic AI Foundation](https://www.linuxfoundation.org/press/linux-foundation-announces-the-formation-of-the-agentic-ai-foundation), a new Linux Foundation body whose platinum members include AWS, Google, Microsoft and OpenAI, so no single vendor owns the standard any more. It is still moving fast: [a major revision](https://modelcontextprotocol.io/specification/2026-07-28) landed on 28 July 2026, and thousands of ready-made connectors now exist. Think of it as USB-C for AI: one plug, and any assistant can talk to any tool that speaks it. Our [AI glossary](/blog/ai-glossary-2026) explains it alongside the other terms you will hear from vendors. Concretely: - **Claude** has a [connectors directory](https://claude.com/connectors) with more than 800 integrations (Google Drive, Gmail, Outlook, SharePoint, Slack, Notion, Jira and so on), plus support for custom connectors on every plan. - **ChatGPT** offers a similar catalog, now called apps (Drive, SharePoint, Teams, HubSpot, Salesforce…). Custom MCP apps that can write as well as read are reserved for its Business, Enterprise and Edu plans; on Plus and Pro, developer mode only lets them read. - And for the systems that matter most to you, such as your ERP, your industry software or your internal database, a developer can build a **custom MCP connector** that both assistants can use. Anthropic’s [**MCP tunnels**](https://claude.com/docs/connectors/mcp-tunnels/overview), a research preview available on request to Claude Enterprise organizations and to agents built on its developer platform, let Claude reach such a connector without exposing it to the public internet, which removes a genuine blocker for software that only runs inside your network. #### What it looks like in practice Once the assistant is connected, questions your team asks all day become one-liners: “Summarize the open tickets for project Y.” “Find the latest signed version of the Durand contract.” “What’s our standard rate for X, and when did we last raise it?” The assistant reads the connected source, answers with references, and your team stops playing detective across six tabs. Our client [Oskar Architecten](/projects/oskar-internal-chatbot) made its scattered project knowledge searchable in one conversation, through a dedicated internal assistant backed by MCP servers we built. #### The three questions to settle first ##### Is it safe to connect ChatGPT or Claude to company data? A connected assistant sees everything the connected account sees, and it can be manipulated. Security researchers spent 2025 proving the point: _EchoLeak_ showed a single crafted email could make Microsoft 365 Copilot leak internal files; _AgentFlayer_ did it to ChatGPT via a poisoned document in Google Drive; [_ShadowLeak_](https://www.radware.com/blog/threat-intelligence/shadowleak/) abused ChatGPT Deep Research’s Gmail connector to exfiltrate inbox data from the cloud side, invisible to your own defenses. All were patched, but the pattern (an assistant that reads untrusted content _and_ holds private data _and_ can communicate out, what researcher Simon Willison calls the [“lethal trifecta”](https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/)) is structural. The practical rules: connect the minimum, prefer read-only and block outbound links and images in answers (EchoLeak and AgentFlayer both carried data out through a rendered link or image), and let an admin, not each employee, decide which sources get plugged in. Letting an admin decide is also becoming a product feature. [**Enterprise-managed authorization**](https://claude.com/docs/connectors/building/enterprise-managed-auth), a stable extension of the MCP standard, lets an admin approve connectors once in the company identity provider, so employees find them already approved when they sign in, limited to the groups and roles they hold anyway, with no consent screen per person. Claude offers it on its Team and Enterprise plans, but each connector’s vendor and your identity provider have to support it too, and that support still varies, so check before relying on it. Knowing which AI tools reach which data is also the groundwork for [complying with the EU AI Act](/blog/eu-ai-act-compliance-smes). Be clear about what it fixes, though: it settles _who gets which connector_, the governance half of the question. It breaks none of the three rings above. An approved connector reading a poisoned document is still an approved connector reading a poisoned document. ##### Where does your data go? On the business tiers of both vendors, your data is not used to train models. That part is settled. Residency is not, and the two differ in a way worth knowing. Anthropic’s own service [stores data in the US](https://privacy.claude.com) and processes it in the US or globally. Companies that need EU processing usually go through a cloud provider’s EU region, and should check which Claude models are available there. OpenAI goes further and stores customer content at rest in-region, Europe included, though only on ChatGPT Enterprise and Edu and only through a sales process. So if your requirement is European storage inside the assistant itself, only ChatGPT offers it today, and only on its top tier. Consumer plans are a different story. OpenAI trains on **consumer** ChatGPT conversations by default unless the user switches it off, and Claude asks each user to choose whether their chats may be used for training. Either way, the decision sits with the employee, not with you. That makes employees pasting company data into personal accounts a real policy problem, not a theoretical one. If your team is going to use AI with internal data, give them a sanctioned, governed way to do it, which is what [moving from shadow AI to one company platform](/blog/from-chaos-to-control-centralizing-your-companys-ai) comes down to. ##### What does it cost for your team? Assistant subscriptions are priced per seat: roughly $20–25 per user per month for a standard seat on [Claude Team](https://claude.com/pricing) or ChatGPT Business, depending on annual or monthly billing. For a 20-person company that is $4,800–6,000 a year before anyone has built the custom connector to your ERP, and the employees who use it daily subsidize the ones who open it twice a month. #### Connectors, or your own internal assistant? For many teams, off-the-shelf connectors are genuinely enough: if your knowledge lives in Google Drive or SharePoint and your governance needs are simple, a Business/Team subscription with two or three connectors is a fine answer. A dedicated [internal chatbot](/packages/internal-chatbot) becomes the better answer when the three questions above start to bite. It runs in a dedicated environment, or entirely on your own premises, with your data under your control. It answers only from a knowledge base you curate, with guardrails, instead of from whatever a connected account can reach. It connects to your real systems (wikis, CRM, ERP, databases) through integrations we build for you. And it is priced as a flat monthly tier sized to the team rather than per seat, whether your people ask three questions or three hundred. That is what Oskar runs today, and why a governed company assistant often costs less than a pile of individual licences. Off-the-shelf connectors | Dedicated internal chatbot Best when | Knowledge lives in Drive or SharePoint | Knowledge is scattered across systems Time to first answer | An afternoon | 2 to 4 weeks What the AI can reach | Whatever the connected account sees | A knowledge base you curate Connects to ERP or industry software | Custom connector, built by you | Built for you Hosting | Vendor cloud; EU storage only on ChatGPT Enterprise and Edu | Dedicated environment or on-premise Pricing | Per seat, at every headcount | Flat tier sized to the team Governance | Central via your identity provider, on business tiers only | One governed door [See the Internal Chatbot package →](/packages/internal-chatbot) #### What we build for you - **Custom MCP connectors** for your ERP, industry software or internal database, usable from both Claude and ChatGPT. Each server exposes one domain of your data rather than a whole account. For [Oskar Architecten](/projects/oskar-internal-chatbot) we built MCP servers that let the assistant search project databases, building regulations and material specifications, each scoped to one domain. See our [custom projects](/services/custom-projects). - **A dedicated internal assistant** when connectors are not enough: the [Internal Chatbot package](/packages/internal-chatbot), usually live within two to four weeks on your own documents. - **Technology we run ourselves**: our AI helpdesk [TicketFlow](/packages/ticketflow) exposes its own MCP server, so a support team can search tickets, change their status and prepare draft replies from MCP clients such as Claude. [Tell us which system](/contact) you need Claude or ChatGPT to reach, and we will tell you whether a connector, the Internal Chatbot package or a custom MCP server fits. _Facts, plans and prices in this post were checked in September 2026 and move fast. Verify current details with each vendor._ ### Custom Tools vs Enterprise SaaS: Why Building Beats Configuring Source: https://flowful.ai/blog/custom-tools-vs-enterprise-saas/ Language: en Description: Building custom internal tools is now faster than configuring generic SaaS. Why senior developers embrace AI coding, and what it means for your business. You’re paying thousands per month for Salesforce, Zendesk, or HubSpot. Your team uses maybe 20% of the features. You’ve spent weeks configuring workflows that still don’t quite match how your business actually operates. There’s another option now. **Custom internal tools tailored to your business can be built faster and cheaper than you’d spend configuring enterprise software.** #### The Real Cost of Enterprise SaaS Enterprise software promises flexibility. In practice, you get: - Per-seat pricing that scales painfully as your team grows - Features you’ll never use subsidizing features you need - Workflows that force you to adapt your process to the software - Consultants and integrators to customize what should have fit from the start The dirty secret? Most businesses don’t need 80% of what these platforms offer. They need a few core workflows that match exactly how they operate. #### Why Custom Used to Be Out of Reach Building custom software used to mean six-figure budgets and six-month timelines. Only enterprises could afford it. Everyone else made do with SaaS that was “close enough.” That equation has changed. [Chris Gregori](https://www.chrisgregori.dev/opinion/code-is-cheap-now-software-isnt) puts it well: > “Code is cheap now. Software isn’t.” What used to take developers weeks now takes hours. AI coding tools have collapsed the cost of writing code. But maintenance, edge cases, and long-term reliability still require human judgment. **For internal tools, that tradeoff works in your favor.** You control the inputs. You’re the only user. You can update when requirements change. [Theo Browne](https://www.youtube.com/@t3dotgg) asks a question worth sitting with: > “What are some things that you would build if you had more time and knowledge?” A CRM that matches your exact sales process. A support tool that handles your specific workflow. A dashboard showing the three metrics you actually care about, not 42. These tools don’t exist because they’re too specific. No SaaS company would build them. That’s exactly what AI makes possible now. #### AI Makes It Possible, But Not Automatic Antirez, creator of Redis, recently [filed a PR](https://github.com/redis/redis/pull/14661) replacing 3,800 lines of C++ with a minimal C implementation: > “This code was written by Claude Code using Opus 4.5 and tested carefully. The code review was independently performed by Codex GPT 5.2.” AI wrote the code. A different AI reviewed it. A human made the final call. This is the pattern that works: AI as a force multiplier, not a replacement for judgment. As Theo puts it: > “We’re all managers now.” But the demos are misleading. Non-coders watch someone build an app in 10 minutes and assume they can do the same. They can, for prototypes and simple tools. Anything beyond that falls apart without the knowledge to evaluate what the AI actually produced. Junior devs get real benefits: faster iterations, quicker bug fixes, less time stuck on boilerplate. Genuine productivity gains. But the multiplier effect lives with senior devs. They accept more AI-generated code than anyone, not because they’re less careful, but because they know how to direct it. Clear specs. Small chunks. Fast reviews. They have what AI coding rewards: **clarity**, **delegation**, and **orchestration**. The result is production-ready code that’s well-architected and easy to maintain. **What matters now is knowing what to build and clearly describing why.** #### The New Math **Enterprise SaaS:** thousands per month, 20% feature utilization, workflows that don’t quite fit. **Custom internal tool:** one-time build, exactly what you need, no per-seat scaling. The SaaS model made sense when custom software required months of development. AI-assisted development changes that equation. But AI won’t build reliable tools on its own. You need engineers who can translate business requirements into clear specs, review AI output, and maintain systems over time. At Flowful, we’re [senior developers and AI engineers](/about). We build [custom internal tools](/services/custom-projects) that replace expensive enterprise licenses with software that fits how you actually work. Our own support desk, [TicketFlow](/packages/ticketflow), and scheduling tool, [Flowcal](/packages/flowcal), started exactly that way. ### AI Email Automation Explained: How a System Learns to Answer Email Like You Source: https://flowful.ai/blog/email-automation-ai-workflows/ Language: en Description: Inside a production AI email system: extracting tone and FAQs from past conversations, knowledge bases, templates, threads and Gmail, Outlook or SMTP. **Email still takes up to 28% of the knowledge worker’s workweek**, according to the [McKinsey Global Institute](https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-social-economy), and most tools that promise to fix it are autoresponders with better marketing. What we build is different: a system that learns from your past conversations, your documentation, and your reply habits, then drafts replies your team only has to review and send. The language model is the least interesting part of that system. What decides whether a draft gets sent or rewritten is everything around it. Here is how the pieces fit together in production. #### It Starts With Your Past Conversations Before writing a word, a good system reads. During setup it goes through a few hundred of your recent threads and extracts two things. The first is a style guide: how you open, how you sign off, how formal you are, how long your sentences run, how you deliver bad news. It lives in a plain document your team can read and correct. If your support team writes “Hi Tom” and never “Dear Mr. Jansen”, the drafts will too. The second is a Q\&A library. Recurring questions get paired with the answers your team actually gave, which usually comes out to 50 to 100 entries covering pricing, shipping, compatibility, returns, whatever your inbox is made of. You go through the list once, drop the outdated entries, and approve the rest. A model without context writes like a press release. Fed your phrasing and your real past answers, it writes drafts that read like your team on a good day. #### A Knowledge Base It Can Search, Tools It Can Call That Q\&A library becomes the core of the knowledge base, indexed with semantic search rather than keywords. “How much is shipping to Belgium?” and “wat kost verzending naar België?” land on the same entry. Product documentation, price lists, and policy pages get indexed alongside it. Static knowledge has limits, though. No FAQ can answer “where is my order?” For that the system calls tools: an order lookup, a stock check, a calendar, a CRM record. On a real email, the two together look like this: > **Inbound, 9:14 am.** “Hi, I ordered an amplifier last Tuesday, order 84312. It still hasn’t arrived. Where is it?” > > **Draft, 9:14 am.** “Hi Tom, sorry for the wait. Your order left our warehouse yesterday, and the carrier expects delivery tomorrow before 6 pm. Here is the tracking link. Let me know if nothing shows up.” Behind that draft: one FAQ entry about delivery delays, one order lookup, phrasing from the style guide. In front of it: a human who clicks send. The harder question is what happens when nothing matches. If neither the knowledge base nor a tool produces the answer, the email is flagged for a human instead of answered with a guess. The same honesty applies to attachments: it will never pretend to have read a document it could not open. [Audiovolt](/projects/audiovolt), a Dutch car audio retailer, runs on this setup for hundreds of product questions a day. Reviewing a grounded draft takes about 30 seconds; writing the same answer from scratch took 5 minutes. [Mailflow](/projects/mailflow-email-auto-responder-building-an-ai-saas), our AI auto-responder, is a simplified self-serve version of the same approach. #### Templates Still Have a Job Not every reply should be generated. Order confirmations, quotes, onboarding sequences, anything a lawyer or an accountant signed off on: that wording should be fixed, reviewed once, and sent identically every time. The interesting cases are hybrids. A quote request comes in, the template provides the structure and the legal footer, and the model fills in the rest: the right products, the right discount tier, and one paragraph that actually responds to what the customer wrote. Our rule of thumb: if someone approved the exact wording, make it a template. If a human would rephrase it for every customer, let the model draft it. [WingBuddy](/projects/wingbuddy-sales-email-assistant), a sales email assistant we built for a client, runs on that hybrid. Their reps write a rough draft, and the AI reshapes it with the company’s templates and the style of its top performers. > **Want this on your own inbox?** Our [Email Automation package](/packages/email-automation) covers the whole chain: style and FAQ extraction, knowledge base, templates, and deployment on your mailboxes, with human validation built in. **[See what’s included](/packages/email-automation)** #### The Plumbing: Threads, Signatures, and Your Actual Mailbox This is the part demos skip and production depends on. Drafts have to appear inside the existing thread, with the history quoted, from your address. Your team reviews them in the inbox it already works in, not in a separate dashboard. Labels handle triage along the way: drafted and ready for review, needs a human, filtered out. Signatures are managed and validated centrally, so every draft leaves formatted and on-brand. None of this should depend on your provider either. The automation talks to a common mail interface, so the same workflows run on Gmail, on Outlook, or on any mailbox that speaks SMTP. Changing providers does not mean rebuilding the automation, and nobody migrates to a new platform. Attachments go both ways: inbound invoices and forms are read and their data extracted, outgoing drafts can carry brochures, spec sheets, or the quote itself. #### Humans Approve, and Everything Is Logged Default mode is draft, not send. Every reply waits for an approve, an edit, or a reject, a principle we apply to [every AI project we deliver](/blog/reliable-ai). Each run is also logged step by step: which filter matched, what the search retrieved, which tools were called, what the model wrote. When a draft is wrong, you can see exactly why, fix the FAQ entry or the template behind it, and the correction carries over to every future draft. Auto-send has to be earned. Once a category has shown weeks of accurate drafts, you can let it go out on its own. Some clients do. Many keep a human on every send and still recover most of the time, because reviewing takes seconds where writing took minutes. #### Built on Our Own Stack The engine underneath is [Vectoria](/blog/inside-vectoria-one-brain-for-every-ai-package), our in-house retrieval and workflow engine, self-hosted on EU infrastructure. Your past emails, product docs, and pricing stay in a knowledge base you control rather than in a third-party vector database. For clients in finance, healthcare, and legal, that matters as much as the hours saved. Setup takes 2 to 4 weeks, extraction and review included, and our clients typically save 15 to 20 hours a week across a team. **[Explore the Email Automation package](/packages/email-automation)**: transparent pricing, one setup fee, three monthly tiers. Want to see it answer a real email first? [Book a free 30-minute discovery call](/contact), bring a handful of typical messages from your inbox, and we’ll walk through what the system would do with each one. ### EU AI Act Compliance for Belgian SMEs: What You Actually Need to Do Source: https://flowful.ai/blog/eu-ai-act-compliance-smes/ Language: en Description: EU AI Act guide for Belgian SMEs, updated after 2 August 2026: Article 50 rules for chatbots and voice agents, Digital Omnibus delays, compliance checklist. Since 2 August 2026, the EU AI Act’s transparency rules apply. If you’re a Belgian SME with a chatbot on your website or an AI agent answering your phone, they already concern you: people must be told they are talking to an AI. The good news: most small business AI use cases still fall into low-risk categories, and the transparency duty is easy to meet. This post gives you a plain-language walkthrough of the regulation, a practical compliance checklist, and specific resources for Belgian companies. We build AI-powered automation for SMEs every day at Flowful, and we design our systems with compliance in mind from day one. Here’s what we’ve learned. We updated this guide on 14 September 2026, after the Digital Omnibus on AI became law and Article 50 started to apply. > **Disclaimer.** Flowful builds AI automation, we are not lawyers or compliance consultants. This post is general information, not legal advice. It shares what we learned compiling our own AI Act posture, so SMEs can ask informed questions of their legal counsel. It reflects the AI Act as amended by the Digital Omnibus, and the Commission’s Article 50 guidelines, as of 14 September 2026. Verify with a specialist before acting on any specific obligation. #### What Is the EU AI Act? Key dates, as amended by the [Digital Omnibus on AI](https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force) ([Regulation (EU) 2026/1744](https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng), published on 24 July 2026, in force since 27 July 2026): - **2 Feb 2025:** Prohibited practices banned. AI literacy (Article 4) applies. - **2 Aug 2025:** GPAI rules apply, together with the governance and penalty chapters. [Code of Practice](https://digital-strategy.ec.europa.eu/en/policies/ai-code-practice) published 10 July 2025. - **2 Aug 2026:** The AI Act applies in general. The Article 50 transparency obligations (chatbots, voice agents, deepfakes, generated content) are live. - **2 Dec 2026:** Deadline for machine-readable marking by generative AI systems already on the market before 2 August 2026. - **2 Dec 2027:** Annex III high-risk obligations apply (postponed from August 2026). - **2 Aug 2028:** Annex I high-risk obligations apply (postponed from August 2027). For a Belgian SME running a chatbot or a voice agent, the date that matters has already passed: the disclosure duty has applied since 2 August 2026. High-risk obligations land late 2027, but if they concern you, do not wait. --- #### What Changed on 2 August 2026 Two things happened this summer. The Digital Omnibus on AI stopped being a proposal: the Council gave its [final approval on 29 June 2026](https://www.consilium.europa.eu/en/press/press-releases/2026/06/29/artificial-intelligence-council-gives-final-green-light-to-simplify-and-streamline-rules/) and the regulation entered into force on 27 July. Then, on 2 August, Article 50 started to apply. The Commission adopted its [Article 50 guidelines](https://digital-strategy.ec.europa.eu/en/library/guidelines-transparency-obligations-providers-and-deployers-ai-systems) on 20 July 2026, next to a [Code of Practice on marking and labelling AI-generated content](https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content). - **AI that talks to people must say so.** Chatbots, voice agents and AI agents must inform people that they are dealing with an AI, clearly and at the latest at the first interaction ([Article 50](https://artificialintelligenceact.eu/article/50/), paragraphs 1 and 5). The exception for cases where this is “obvious” is narrow: the guidelines name helpdesk chatbots as a case where it does not apply. - **Generated content must be marked.** Providers of systems that generate audio, images, video or text must mark the output in a machine-readable way. Systems already on the market before 2 August 2026 have until **2 December 2026**. That grace period covers marking only: disclosure in conversations had to be in place on 2 August. - **Deepfakes and unreviewed public-interest text need a visible label.** This duty sits with the business that publishes them. Content published before 2 August 2026 does not need retroactive labels, but content generated earlier and published after that date does. - **High-risk dates moved.** Annex III systems (hiring, credit scoring, education and others) now apply from **2 December 2027**, AI in products covered by Annex I from **2 August 2028**. - **Lighter rules for smaller companies.** Some simplifications reserved for SMEs now extend to small mid-caps, and the Article 4 AI literacy duty was softened: businesses must take measures to support their staff’s AI literacy, without guaranteeing a set level, with the Commission and Member States taking a stronger role in promoting it ([Commission summary](https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force)). - **Fines.** A breach of Article 50 falls in the tier of up to 15M EUR or 3% of worldwide annual turnover ([Article 99](https://artificialintelligenceact.eu/article/99/)). For SMEs, the lower of the two is the ceiling. ##### What a Chatbot or Voice Agent Deployer Should Do Now On paper, Article 50(1) is a design duty for the provider: whoever builds the system or puts it into service under their own name. That is us when you run a Flowful package, and it is you if your team built the bot in-house. Either way, your customers see your brand, so check these six points yourself. 1. **Disclose in the first message.** Not only in the terms and conditions or a linked document: the guidelines call that insufficient on its own. For example: “Hello, I’m the AI assistant of \[Company]. I can answer questions about quotes and opening hours, or put you through to a colleague.” A bare “Assistant” label, or a line like “this service uses LLMs”, does not count. 1. **Say it at the start of every call.** A voice agent should state that it is an AI in its greeting, before the caller explains anything: “Hello, you’ve reached \[Company]. You’re speaking with an AI assistant. I can book an appointment or take a message.” On longer calls, the guidelines recommend reminders. A tone or jingle alone is not enough. 1. **Answer honestly when asked.** If someone asks “am I talking to a real person?”, the system has to say it is an AI. The same applies when the conversation shows the person is confused about it. 1. **Keep the disclosure accessible.** Plain words, readable by screen readers, simpler still if children or elderly people are likely users. In sensitive flows such as complaints, insurance, health, legal or financial questions, repeat it during the conversation. 1. **Label what you publish.** AI images or video of realistic-looking people, cloned voices, and AI-written texts on matters of public interest that nobody reviewed editorially need a visible label. 1. **Ask your vendors about marking.** If a tool generates images, audio, video or text for you, its provider must mark that output in a machine-readable way, by 2 December 2026 for tools already on the market before 2 August. Get that confirmed in writing. A route to a human is not an Article 50 requirement. We still build one in by default, because a disclosure that leads nowhere frustrates customers. --- #### The Four Risk Categories The AI Act sorts systems into four tiers, shown in the pyramid above. The higher up, the stricter the rules. Below, what each tier means for an SME. ##### Unacceptable Risk (Banned) Social scoring, subliminal manipulation, real-time biometric identification in public spaces (with narrow law-enforcement exceptions), emotion recognition at work or school, untargeted facial-image scraping. **For SMEs:** very unlikely you are doing any of this. But verify any tool that claims emotion detection, trustworthiness scoring or facial recognition, and get legal advice if it does. ##### High Risk (Heavy Obligations) Listed in [Annex III](https://artificialintelligenceact.eu/annex/3/): HR/recruitment screening, credit scoring, life or health insurance risk and pricing, education access, essential public services, law enforcement, migration, non-real-time biometrics. Plus AI safety components in regulated products under Annex I (medical devices, vehicles, machinery). **For SMEs:** if you screen job applicants, assess loan eligibility, or take consequential decisions about individuals, you are likely here. Chapter III obligations (risk management, data governance, technical docs, human oversight, conformity assessment, EU database registration) apply by **2 Dec 2027** for Annex III and **2 Aug 2028** for Annex I. ##### Limited Risk (Transparency Obligations) Customer-facing chatbots, virtual assistants, deepfakes and certain AI-generated text on matters of public interest, plus emotion recognition or biometric categorisation where not banned. **For SMEs:** since 2 August 2026, tell users from the start that they are talking to an AI, and label deepfakes or unreviewed AI-generated text published to inform the public. Article 50 obligation, straightforward to implement. ##### Minimal Risk (No Specific Obligations) Spam filters, AI-assisted email drafting, workflow automation, product recommendations, inventory forecasting, document classification, AI-assisted translation. **For SMEs:** almost everything you use daily. No mandatory steps. Document what you run and why. ##### A Separate Track: General-Purpose AI (ChatGPT, Claude, Gemini) General-purpose AI models (GPAI) sit alongside the four risk tiers under their own regime, applicable since 2 August 2025. Providers of these models must publish a summary of training data, respect EU copyright (notably the [Article 4(3) text-and-data-mining opt-out](https://eur-lex.europa.eu/eli/dir/2019/790/oj)), and supply technical documentation to downstream users. The largest models (above 10²⁵ FLOPs of training compute) face additional systemic-risk obligations. **For SME deployers using ChatGPT, Claude or Gemini in a workflow:** the practical impact is light. Keep using them, apply the Article 50 transparency rules (disclose AI to users, label deepfakes and unreviewed public-interest text), and record which model handles what in your AI inventory. --- #### What This Means for Belgian SMEs If you use AI chatbots for support, email automation, workflow tools or document processing, your use cases are almost certainly **minimal or limited risk**. You are probably fine. But “probably” is not a strategy. Four reasons to still pay attention: 1. **You might be high-risk without realising it.** HR screening CVs, sales scoring leads feeding into credit decisions, anything consequential about individuals: the classification depends on the use case, not the technology. 1. **Provider or deployer, the role decides the obligations.** The AI Act splits responsibility between **providers** (who develop or place an AI system on the market) and **deployers** (who use it under their own authority in a professional context). Most SMEs are deployers, sometimes both at once. Providers carry the bulk of high-risk and GPAI obligations; deployers must follow the provider’s instructions for use, monitor operation, keep logs, and in high-risk contexts run a fundamental-rights impact assessment under [Article 27](https://artificialintelligenceact.eu/article/27/). Deploy a general-purpose tool in a high-risk context and the burden lands on you. 1. **Belgian enforcement is still being set up.** The federal government agreement names [BIPT](https://www.bipt.be/operators/digital/ia-act/application-of-the-ai-act) as the main market surveillance authority, [SPF Economie](https://economie.fgov.be/fr/themes/entreprises/ai-act) coordinates implementation, and the CSA is one of twenty-one Article 77 fundamental-rights bodies (audiovisual media in the French-speaking community). Belgium missed the August 2025 deadline to designate its authorities, and as of mid-September 2026 we could not find a Belgian law that formally does so. The regulation applies directly all the same. In France, the bill giving the CNIL supervisory powers under the AI Act passed the Senate on 18 February 2026 and was still [pending before the Assemblée nationale](https://www.assemblee-nationale.fr/dyn/17/dossiers/DLR5L17N53140) when we updated this post. Fines under [Article 99](https://artificialintelligenceact.eu/article/99/): up to **35M EUR or 7%** of global turnover (prohibited practices), **15M / 3%** (most other obligations, including Article 50 transparency), **7.5M / 1%** (false information). SMEs and start-ups pay up to the **lower** of the two figures, still significant. 1. **Clients will start asking.** B2B buyers, especially larger companies and public sector, will ask about your AI compliance posture. Being prepared is an edge. --- #### A Practical Compliance Checklist Seven steps you can start today. These are baseline good practices we follow ourselves, not a substitute for a formal conformity assessment. No law firm on retainer required. 1. **Inventory your AI systems.** List every tool you run, including third-party SaaS. Note what it does, what data it processes, who is affected, and who provides it. You cannot assess risk on what you do not know. 1. **Classify each system.** Three questions: does it influence consequential decisions about people, does it interact with users who may not know they are dealing with AI, can it manipulate or exploit vulnerabilities? Three “no” puts you in minimal or limited risk. One “yes”, check [Annex III](https://artificialintelligenceact.eu/annex/3/) and dig deeper. 1. **Implement transparency.** Disclose AI to users at the first interaction: in the chatbot’s first message, in the phone greeting. Label deepfakes, and AI-generated text published to inform the public without editorial review. Article 50 obligation, in force since 2 August 2026 and detailed above. A route to a human is not required, but it is good practice. Illustrative visuals that do not resemble real people, places or events need no visible label. 1. **Keep humans in the loop.** Review AI output before it is sent or used for decisions. Build escalation paths. Let employees override. Mandatory and technical for high-risk; good practice everywhere else. For the engineering side of making AI outputs trustworthy enough to act on, see our note on [building reliable AI systems](/blog/reliable-ai). 1. **Document everything.** Inventory, risk classification, justification, transparency and oversight measures, incidents, GDPR data-processing records. Documentation is the backbone of compliance. The [Future of Life Institute compliance checker](https://artificialintelligenceact.eu/assessment/eu-ai-act-compliance-checker/) is a useful free starting point. 1. **Review vendor contracts.** Does the provider classify their system’s risk level? Do they provide the AI Act technical documentation? Who owns the conformity assessment? What happens to your data? Where is it processed? A vendor that cannot answer is a red flag. 1. **Train your team.** The [Article 4 AI literacy](https://artificialintelligenceact.eu/article/4/) obligation has applied since 2 Feb 2025. Since the Digital Omnibus, it asks you to take measures that support your staff’s AI literacy rather than guarantee a level. In practice: make sure staff know what AI they use, how it works at a basic level, its limits, and your internal policy. PhD-level not required. --- #### GDPR and the EU AI Act Work Together If you are already GDPR-compliant, most of the analytical work maps over: **DPIAs** ≈ AI Act risk assessments. **Data minimisation and purpose limitation** ≈ AI Act data governance. **Article 22** ≈ AI Act transparency obligations. **GDPR right to human intervention** ≈ AI Act human oversight. Where the AI Act goes further: technical standards on the system itself (accuracy, robustness, cybersecurity, technical documentation). GDPR governs data; the AI Act governs the system. Do not run two parallel projects. Integrate AI Act work into your existing GDPR framework, same team, same docs. --- #### How Flowful Approaches AI Compliance At [Flowful](/about), we build web and internal chatbots, AI phone receptionists, and email automation for SMEs in Belgium and France, plus an AI-first helpdesk (TicketFlow) and AI-ready booking (Flowcal). We design our systems with compliance in mind from day one. **We are not a law firm.** What follows describes how we build, not legal advice. - **EU-first hosting.** Workflow infrastructure runs on Hetzner (Germany). AI inference, voice, and transactional email route through a small set of carefully selected sub-processors under DPAs with appropriate transfer safeguards (SCCs or technical controls such as no-training, no-retention). On request, we restrict processing to EU-only providers or run open-source models on dedicated infrastructure. The current sub-processor list is in our DPA. - **No training on your data.** Every sub-processor is configured to disable training on customer data; per-vendor settings are documented in our DPA. - **Human-in-the-loop where it matters.** Email Automation can be configured with human approval before sending, and we recommend it for outbound or higher-stakes flows. Voice agents escalate to a person when out of scope. - **AI Act tier by package.** Our **[Web Chatbot](/packages/web-chatbot)**, **[AI Phone Receptionist](/packages/ai-phone-receptionist)** and **Internal Chatbot** sit in Limited Risk. Each conversation opens with a clear AI disclosure, written in the chatbot’s first message and spoken in the receptionist’s greeting, which is what Article 50 asks for. Each also offers a route to a human, which Article 50 does not require but your customers will expect. Our **Email Automation** sits in Minimal Risk. We can add human approval before sending depending on the use case. We do not build Annex III high-risk systems (hiring, credit scoring, insurance pricing, education access) without a formal compliance plan. - **DPA available on request.** Covers GDPR and AI-specific obligations including the no-training clause and the sub-processor list. [Get in touch](/contact). --- #### Belgian-Specific Resources ##### Regulatory and Government - **[BIPT (IBPT)](https://www.bipt.be/operators/digital/ia-act/application-of-the-ai-act):** Belgium’s main market surveillance authority for the AI Act. This is where enforcement will happen for most providers and deployers, including a single point of contact for high-risk system operators. - **[SPF Economie (FPS Economy)](https://economie.fgov.be/fr/themes/entreprises/ai-act):** Coordinates Belgium’s implementation of the AI Act. The dedicated AI Act section has guidance for entrepreneurs, plus an SME-oriented campaign and a downloadable guide. - **[Data Protection Authority (APD/GBA)](https://www.dataprotectionauthority.be/):** Belgium’s GDPR supervisor. As AI compliance and data protection overlap significantly, the APD remains relevant for AI-related data processing questions. - **[AI4Belgium Coalition](https://ai4belgium.be/):** A multi-stakeholder initiative that published Belgium’s AI strategy. Their resources include practical guidelines and sectoral recommendations. ##### Regional Innovation Support - **[Innoviris](https://www.innoviris.brussels/) (Brussels):** Brussels’ innovation funding body. Offers AI-specific vouchers and funding programs. If you’re a Brussels-based SME exploring AI, they can help fund a compliant implementation from the start. - **[Digital Wallonia](https://www.digitalwallonia.be/) (Wallonia):** Wallonia’s digital strategy hub. Runs the Start IA and Tremplin IA programs, and publishes practical AI adoption guides. - **[VLAIO](https://www.vlaio.be/) (Flanders):** Flanders’ innovation and entrepreneurship agency. Offers R\&D grants and SME support programs applicable to AI projects. ##### Practical Tools - **[EU AI Act Explorer](https://artificialintelligenceact.eu/):** An independent resource with a searchable, annotated version of the full regulation. Useful for looking up specific articles and requirements. - **[Future of Life Institute AI Act Compliance Checker](https://artificialintelligenceact.eu/assessment/eu-ai-act-compliance-checker/):** A free self-assessment tool that helps you determine whether your AI system might be high-risk under the AI Act. We’ve also published a [guide to funding AI projects in Belgium](/blog/fund-ai-project-belgium), which covers subsidies and grants that can help offset the cost of building AI solutions that are compliant from the start. --- #### What to Do Next A realistic timeline for a Belgian SME: - **Now:** check that every chatbot and voice agent says it is an AI from the first exchange, label deepfakes and unreviewed public-interest text, inventory your systems, verify no prohibited practice, review vendor contracts. - **By 2 December 2026:** get confirmation from the providers of your generative tools that their output is marked in a machine-readable way. - **Before 2 December 2027:** if a system falls under Annex III, finish classification and high-risk remediation. - **Ongoing:** follow Commission guidance, the formal designation of Belgian authorities, and harmonised standards. Keep documentation current. For most Belgian SMEs running standard business automation, the workload is manageable. Start with the disclosure, classify honestly, build the documentation habit. This post is general information, not legal advice or a compliance assessment. For a formal conformity check, work with a legal or compliance specialist. If you want a [web chatbot](/packages/web-chatbot) or an [AI phone receptionist](/packages/ai-phone-receptionist) that discloses itself properly from the first message, both packages ship with it built in. [Get in touch](/contact) to set one up. ### From Chaos to Control: Centralizing Your Company’s AI Source: https://flowful.ai/blog/from-chaos-to-control-centralizing-your-companys-ai/ Language: en Description: Replace scattered ChatGPT accounts with a secure, centralized internal chatbot. Allowing employees to manage their own ChatGPT subscriptions offers immediate convenience, but it eventually leads to “Shadow AI.” Very quickly, you face scattered costs, zero visibility into data security, and a lack of standardization. Perhaps most critically, the answers your teams receive aren’t aligned with your company’s internal data or context. A cleaner, more scalable approach is to establish an **[internal chatbot](/packages/internal-chatbot)**: a single AI platform for the entire organization. #### **1. Consolidating the Subscription Jungle** Instead of juggling dozens of individual accounts, an internal chatbot provides one interface that connects to LLMs via API. This shift offers immediate benefits: - **Cost Efficiency:** You pay for actual usage, not idle subscriptions. - **Data Sovereignty:** You decide where conversations are stored and what data hits the models. - **Vendor Flexibility:** You avoid lock-in by plugging in OpenAI, Anthropic, Gemini, or open-source models behind the same interface. Tools like **LibreChat**, **Open WebUI**, or **Anything LLM** offer a solid foundation for this. They provide a familiar, multi-user environment, complete with file uploads and history—that runs securely on your own domain. #### **2. Structuring Access and Governance** An internal chatbot acts as the official gateway to AI, putting an end to “DIY AI.” By integrating with your existing SSO, you can automate onboarding and ensure employees only access the tools relevant to their roles. - **Centralized Control:** IT can manage permissions, monitor budgets, and reset accounts easily. - **Role-Based Access:** Create custom agents for specific departments (e.g., HR, Sales, IT) and limit availability to those who need them. #### **3. Unlocking Value with Connected Agents** The true power of an internal platform emerges when it stops just “chatting” and starts working with your data. You can deploy specialized agents connected to your internal systems via RAG (Retrieval-Augmented Generation) and APIs. If your team would rather reach those systems from Claude or ChatGPT directly, [here is how MCP connectors work and what to settle first](/blog/connect-claude-chatgpt-internal-tools). - **Internal Support Agent:** Connected to your knowledge base to resolve [ticketing issues](/packages/ticketflow). - **HR Agent:** Trained on internal policies to answer questions about benefits and procedures. - **Sales Agent:** Integrated with your CRM to draft emails and summarize account details. #### **Getting Started** Moving from scattered accounts to a unified platform begins with auditing your current tools and use cases. At **[Flowful AI](/about)**, we help companies design and deploy internal chatbots built on open-source blocks. We ensure your teams keep the ChatGPT experience they love, while you regain control over costs, data, and infrastructure. ### How to Fund Your AI Project in Belgium: Subsidies, Grants, and Tax Credits Source: https://flowful.ai/blog/fund-ai-project-belgium/ Language: en Description: Belgian funding for AI and digital projects, checked September 2026: Prime Digitalisation, Chèques-Entreprises, Start IA, Tremplin IA, VLAIO, tax incentives. Most Belgian SMEs don’t know they can get roughly 15 to 75% of their AI project funded by the government. We see it all the time: business owners assume subsidies are reserved for universities, pharma labs, or companies with dedicated grant writers on staff. They’re not. If you’re a small or medium-sized business in Belgium exploring AI, there’s a good chance public money is available to cover a significant chunk of your costs. The catch? The landscape is fragmented. Each region has its own programs, its own rules, and its own application process. Federal tax incentives add another layer. It’s not complicated once you understand the map, but most companies never get that far. They either don’t know the programs exist, or they start looking and give up because the information is scattered across a dozen websites in three languages. This post is the map. We’ll walk through every major funding option by region, explain what each one covers, and share practical tips on what actually gets applications approved. **Checked in September 2026.** Since this guide first ran in February, Innoviris has suspended its programmes for 2026 and Digital Wallonia has relaunched Start IA and Tremplin IA at new rates. In Brussels, the open route for digital work is now the Prime Digitalisation. Here is where each programme stands: Programme | Region | Status in September 2026 | Support Prime Digitalisation | Brussels | Open, apply before the mission starts | 25 to 70%, max 10,000 EUR per calendar year Start AI, GENAI, Proof of Business (Innoviris) | Brussels | Suspended for 2026 | No new calls Start IA | Wallonia | Call closes 30 September 2026 | Up to 50%, max 5,000 EUR Tremplin IA 08 | Wallonia | 2026 call published 10 June 2026, check the deadline | Up to 50%, max 20,000 EUR Chèque maturité numérique | Wallonia | Open | 50%, max 50,000 EUR over 3 years Chèque cybersécurité | Wallonia | Open | 75%, max 50,000 EUR over 3 years KMO-portefeuille | Flanders | Training, and cybersecurity advice only. Confirm on vlaio.be | 20 to 45%, max 7,500 EUR per year Technology deduction | Federal | In force | 13.5% one-off or 20.5% spread --- #### Brussels-Capital Region Two funders matter in Brussels. [Brussels Economy and Employment](https://economie-emploi.brussels/prime-digitalisation) runs the Prime Digitalisation, which is open. [Innoviris](https://www.innoviris.brussels/), the regional research and innovation agency, has suspended its programmes for 2026. ##### Prime Digitalisation (Digitalisation grant) **Best for:** Brussels SMEs hiring an outside provider to digitalise a process, secure their IT, or rebuild their website. The [Prime Digitalisation](https://economie-emploi.brussels/prime-digitalisation) pays part of a consulting mission in one of three areas: - the digitalisation of your internal processes, your means of production, or your products and services - the IT security of your company - the technical development or improvement of your website **Who is eligible:** SMEs with at least one operating site in the Brussels-Capital Region, active in one of the eligible sectors listed on the official page. The mission must solve a one-off problem your team lacks the skills for, last 6 months at most, and cannot be permanent subcontracting. You also commit to the region’s responsible digital charter. **Rate:** 25% of eligible costs by default. Bonuses raise it: +25% for a starter (registered for less than 4 years) that is a micro or small company, +20% for a medium one, and +30% (micro or small) or +20% (medium) for each recognised environmental or social exemplarity. The total is capped at **70%**. **Ceiling:** **10,000 EUR per company per calendar year**, with a minimum grant of 500 EUR per mission and at most two funded missions a year. **The provider:** consulting in the field must be their main activity, for at least 2 years. They must be independent from your company and invoice you, directly or through a billing service. **Apply before work starts.** The request goes through MonBEE at the latest the day before the mission begins, and the mission can start at the earliest the day after you file. A mission that has already started cannot be funded. An automation or chatbot project can fit the first area when it digitalises a specific process, such as handling inbound emails or answering customer questions. Brussels Economy and Employment decides case by case, so describe the process the mission changes and check eligibility before you sign. hub.brussels also explains the grant in its [entrepreneur guide](https://info.hub.brussels/en/guide/subsidies-entrepreneurs/digitalisation-grant). ##### Innoviris programmes: suspended in 2026 Innoviris states on its programme pages that, given its 2026 budget, **no new calls will be launched and no new projects funded in 2026**. Projects with a signed agreement continue. Applications already submitted are still processed and may be funded in 2027, depending on available resources. This covers the programmes the earlier version of this guide described: - [Start AI innovation voucher](https://www.innoviris.brussels/program/innovation-vouchers-start-ai): up to 75% of expenses excluding VAT, capped at 10,000 EUR per calendar year. - [GENAI](https://www.innoviris.brussels/program/genai): generative AI feasibility study and proof of concept, budgets up to 80,000 EUR excluding VAT. Its last cut-off was 13 June 2025. - [Proof of Business](https://www.innoviris.brussels/program/proof-business): 50 to 70% of project costs. Watch [innoviris.brussels](https://www.innoviris.brussels/) for the 2027 calls. [SustAIn.brussels](https://sustain.brussels/), the Brussels Digital Innovation Hub, still offers supplier matchmaking, training, and proof of concept support. Ask them directly what is free in 2026. ##### Practical tips for Brussels applications - **File before you fix a start date.** The Prime Digitalisation excludes missions that began before the request. Put the filing date in the provider’s schedule. - **Check the provider’s profile.** A provider whose main activity is not consulting, or with less than 2 years in the field, makes the mission ineligible. - **Keep R\&D projects in view for 2027.** An application already filed with Innoviris may still be funded in 2027, depending on available resources. --- #### Wallonia Wallonia funds AI through two [Digital Wallonia](https://www.digitalwallonia.be/ia/) calls, Start IA to find use cases and Tremplin IA for proofs of concept, and funds wider digital work through the [Chèques-Entreprises](https://www.cheques-entreprises.be/). ##### Start IA (Digital Wallonia) **Best for:** Walloon SMEs that want an expert to find and rank their AI use cases. **Deadline: the 2026 call runs from 10 June to 30 September 2026.** [Start IA](https://www.digitalwallonia.be/fr/publications/start-ia/) funds an AI expert mission of up to 4 months: 5 mandatory days for the audit and action plan, plus up to 5 optional days for advanced data analysis, process optimisation, or building an AI agent. The expert identifies use cases, assesses their feasibility and impact, and structures a roadmap. **Who is eligible:** private SMEs (fewer than 250 employees, turnover up to 50 million EUR or balance sheet up to 43 million EUR) with economic activity in Wallonia and a company number. **Rate and ceiling:** Digital Wallonia funds up to **50%** of costs, with a maximum public contribution of **5,000 EUR** per participant. The call has a 500,000 EUR budget for more than 100 projects. This replaces the format described in the earlier version of this guide (45 hours at 70%, open to organisations of any size). ##### Tremplin IA (Digital Wallonia) **Best for:** Walloon SMEs ready to build an AI proof of concept with a provider. [Tremplin IA](https://www.digitalwallonia.be/fr/publications/tremplin-ia/) is in its eighth edition, published on 10 June 2026. It covers up to **50% of the provider’s fees**, capped at **20,000 EUR** per project, from a 1 million EUR budget for more than 50 projects. Eligibility uses the same SME definition as Start IA. You apply together with your AI provider: an application of at least 7 pages on the official template, sent through the online form, with a signed de minimis aid declaration. The call page does not state a closing date, so check it before you plan a project around this call. ##### Chèque maturité numérique (Chèques-Entreprises) **Best for:** Walloon companies that want paid support to assess and digitalise their processes, including the process an AI tool would change. The [chèque maturité numérique](https://www.cheques-entreprises.be/cheques/maturite-numerique) pays part of a labelled provider’s work to assess your digital maturity and plan the digitalisation of your processes: infrastructure, information flows, production processes, and the organisation of work, up to the specifications and the follow-up of implementation. **Who is eligible:** companies, through the Chèques-Entreprises platform. **Rate:** **50%** of the cost. **Ceiling:** **50,000 EUR excluding VAT over 3 years**. The provider must be labelled on the platform. For a first request, the provider opens the file, and you pay your share before the administration accepts it. The programme page does not mention AI by name. Read the full conditions in the procedure on wallonie.be, linked from the programme page, before the mission starts. ##### Chèque cybersécurité The [chèque cybersécurité](https://www.cheques-entreprises.be/cheques/cybersecurite) covers **75%** of an audit or diagnostic of your cybersecurity followed by the implementation of the recommended actions, up to **50,000 EUR excluding VAT over 3 years**. It is worth considering before any project that connects an AI tool to your email, CRM, or customer data. ##### Practical tips for Wallonia applications - **Start IA closes on 30 September 2026.** Prepare your application now. - **Write Tremplin IA with your provider.** The 7-page application is a joint document, so pick a provider who has written one before. - **Think in steps.** Start IA to choose the use case, Tremplin IA for the proof of concept, then a chèque maturité numérique for the wider rollout. Check each call’s rules on combining aid: each cost can only be funded once. --- #### Flanders Flanders channels most of its innovation support through [VLAIO](https://www.vlaio.be/en/subsidies) (Flanders Innovation & Entrepreneurship). The programs here are broader than AI-specific, but they absolutely cover AI projects. **Before you apply:** the Flanders figures below could not be re-checked in September 2026. Confirm them on [vlaio.be](https://www.vlaio.be/). ##### VLAIO Research and Development Projects **Best for:** Companies building innovative AI products or services with a genuine R\&D component. VLAIO offers separate tracks for [research projects](https://www.vlaio.be/en/subsidies/research-project) (longer-term knowledge building) and [development projects](https://www.vlaio.be/en/subsidies/development-project) (shorter-term innovation leading to new products, processes, or services). The rates differ by track: - **Development projects:** base rate of **25%** (regardless of company size), minimum support €25K - **Research projects:** base rate of **25% (large)**, **35% (medium)**, or **45% (small)** enterprises, minimum support €100K - **+10%** for collaboration with another independent company (at least one SME) - **+15%** for international or interregional collaboration - Total funding can reach **up to 50-60%** for development and **up to 70%** for research, depending on your setup These are substantial grants for serious projects. VLAIO also offers [R\&D feasibility studies](https://www.vlaio.be/en/subsidies/rd-feasibility-study) if you need to validate your approach before committing to a full project. ##### KMO-Portefeuille (SME E-Wallet) **Best for:** Training your team on AI tools or getting cybersecurity consulting. The [KMO-portefeuille](https://www.vlaio.be/en/subsidies/sme-e-wallet) subsidizes training and consulting from registered service providers: - **Small enterprises:** 30% subsidy (45% for cybersecurity) - **Medium enterprises:** 20% subsidy (35% for cybersecurity) - **Maximum:** 7,500 EUR per year Important note: since February 1, 2026, the advisory component of KMO-portefeuille is limited to **cybersecurity only**. Digitalization advisory is no longer supported. Training services (including for digitalization and AI topics) remain available for all themes. ##### Practical tips for Flanders applications - **Frame your AI project as innovation, not just technology adoption.** VLAIO R\&D grants reward novelty. Explain what’s new about your approach, not just that you’re using AI. - **Collaboration boosts your rate.** If you can partner with another SME or a research institution, your subsidy percentage goes up significantly. - **Start with KMO-portefeuille for quick wins.** It’s the fastest program to access and can fund AI training for your team. For advisory work, focus on the cybersecurity angle if applicable. --- #### Federal Tax Incentives (All Regions) Regardless of where your company is based, Belgian federal tax incentives can stack on top of regional subsidies. These apply to your corporate tax return and can meaningfully reduce the cost of AI investment over time. ##### Technology Deduction Since the January 1, 2025 tax reform, the former “investment deduction for R\&D” has been renamed the **technology deduction** (_technologie-aftrek / déduction pour technologie_). It applies specifically to **patents** and **environmentally friendly R\&D investments**, not general software development or AI projects. If your AI work leads to a patent or qualifies as an environmentally friendly R\&D investment, you can claim a deduction of **13.5%** of the acquisition value as a one-shot deduction, or **20.5%** if you spread it over the depreciation period. This directly reduces your taxable income. Alternatively, you can opt for a tax credit calculated at 25% of those rates. If you have insufficient taxable profits to use the deduction, unused tax credits are automatically reimbursed after four years. ##### Investment Deduction for Digital Assets Small companies (and natural persons) also get a basic investment deduction on fixed assets acquired from 1 January 2025: 10% in general and **20% for digital fixed assets**. Ask your accountant which software and hardware in an AI project qualify. The [FPS Finance page](https://finances.belgium.be/fr/entreprises/impot_des_societes/avantages_fiscaux/deduction_pour_investissement) lists the current rates (in French). ##### Innovation Income Deduction Companies generating revenue from qualifying intellectual property (including copyrighted software) can deduct **85% of the net IP income** from their taxable base. This effectively lowers the tax rate on that income to around **3.75%** instead of the standard 25% corporate tax rate. This applies to patents, copyrighted software, plant breeders’ rights, and certain other IP. If you build an AI product and license or sell it, the revenue can qualify, provided the software results from a qualifying R\&D program and passes the nexus test (the deduction is proportional to your own R\&D spend versus acquired IP). It’s not automatic for all software revenue. New rules from 2025 also allow unused deductions to be carried forward as a non-refundable tax credit. ##### Partial Exemption of Withholding Tax for Researchers If you employ R\&D staff (developers, data scientists, AI engineers) who hold qualifying degrees (master’s, PhD, or qualifying bachelor’s degrees in sciences, engineering, or related fields), your company can be exempted from paying up to **80% of the wage withholding tax** for those employees. Holders of qualifying bachelor’s degrees are eligible at the 80% rate, but capped at 25% of the total master’s/PhD exemption amount (50% for SMEs). The exemption is proportional to the time they spend on R\&D activities. This is one of the most impactful incentives for companies building AI in-house. A developer spending 80% of their time on R\&D effectively costs you significantly less in payroll taxes. You do need to register your R\&D projects with [BELSPO](https://www.belspo.be/) (the Belgian Science Policy Office) before starting, and maintain proper time-tracking records. For detailed guidance on all federal incentives, the [FPS Finance investment deduction page](https://finances.belgium.be/fr/entreprises/impot_des_societes/avantages_fiscaux/deduction_pour_investissement), the [PwC Belgium tax summary](https://taxsummaries.pwc.com/belgium/corporate/tax-credits-and-incentives) (reviewed September 2026), and [BELSPO](https://www.belspo.be/) (for WHT exemption registration rules) are reliable references. --- #### Combining Programs for Maximum Impact Here’s what many companies miss: you can combine regional subsidies with federal tax incentives. A small Brussels company could use the Prime Digitalisation for a mission that automates a process, then claim the 20% deduction on digital fixed assets it paid for itself. A Flemish company could use KMO-portefeuille for AI training, then apply for a VLAIO development project for the build, while claiming the withholding tax exemption for its R\&D developers the entire time. A few important constraints to keep in mind: - You can’t “double-fund” the same costs. Each euro of expenditure can only be covered by one subsidy. - The technology deduction and the R\&D tax credit are **mutually exclusive** per asset: you must choose one or the other. - Different cost categories within the same project often qualify for different programs, so combining is still very effective. --- #### Do you already have an AI project in mind? An [email automation](/packages/email-automation) or [web chatbot](/packages/web-chatbot) package has a fixed setup fee and a defined scope, so its quote is easy to compare with a programme’s ceiling. For larger work, our [proofs of concept](/blog/proof-of-concept-the-smart-way-to-start-your-ai-project) and [our projects](/projects) show what [Flowful AI](/about) builds for Belgian businesses. **[Compare our packages](/packages)** or [book a call](/contact) to see which programme fits your project. We can point out which subsidies are likely to apply. The application itself is best handled by you directly, or through a specialised firm. ### How to Fund Your AI Project in France: Grants, Subsidies, and Tax Credits Source: https://flowful.ai/blog/fund-ai-project-france/ Language: en Description: A practical guide to French funding for AI projects. CIR, CII, BPI France, France 2030, regional grants, and how to get your application approved. France is investing heavily in artificial intelligence. With the France 2030 plan, billions of euros mobilized by BPI France, and tax incentives like the Crédit d’Impôt Recherche, French businesses have a wide array of funding options for AI projects. Yet many SMEs and mid-caps miss out, either because they don’t know the programs exist or because the administrative complexity puts them off. This post is your roadmap. We cover every major funding mechanism and explain what each one covers in practice. --- #### Tax credits: CIR and CII ##### Crédit d’Impôt Recherche (CIR) The CIR is France’s flagship R\&D tax credit, and AI projects are fully eligible. **What it offers:** - **30%** tax credit on R\&D expenses (up to 100 million euros per year) - **5%** beyond 100 million - **50%** for companies in overseas territories **Eligible expenses:** - Salaries of researchers and research technicians (gross + employer contributions) - Operating costs (flat rate of 40% of personnel costs, rate in effect since February 2025) - Subcontracting to approved research organizations - Equipment depreciation dedicated to R\&D **Important:** the CIR does not cover simple deployment of existing AI tools. Your project must involve genuine scientific or technical uncertainty. Configuring ChatGPT for your customer service is not eligible. Developing a custom RAG system integrating your proprietary business data, however, may well be. ##### Crédit d’Impôt Innovation (CII) Reserved for SMEs, the CII covers downstream innovation phases. - **20%** of eligible expenses (rate in effect since January 2025) - Cap of 400,000 euros in expenses, yielding a maximum credit of **80,000 euros per year** - Higher rates in Corsica (35% for medium enterprises, 40% for small ones) - **Eligibility:** SMEs only (under 250 employees, turnover under 50 million euros) - **What’s covered:** design of prototypes or pilot installations of new products with superior performance compared to existing products - **Validity:** until December 31, 2027 --- #### BPI France: loans, grants, and diagnostics BPI France offers a range of support for AI projects. ##### Aide pour le Développement de l’Innovation (ADI) - **Up to 2 million euros** (grant + repayable advance) - Covers up to **45%** of projected expenses - Interest rate: 0%, repayment deferral up to 36 months ##### Prêt Innovation R\&D - 50,000 to **3 million euros** (“Classique” variant) - 5-8 year term with 1-3 year deferral - For independent SMEs and mid-caps ##### Subvention Innovation - Up to **50,000 euros** (up to 70% of expenses) for innovative SME projects ##### IA Booster France 2030 A program to accelerate AI adoption: - Covers up to **80%** of diagnostic and advisory costs (initial phases), 50% for implementation support - Target: SMEs and mid-caps with 10 to 2,000 employees - Includes: self-assessment, online training, personalized advisory ##### Diag Data IA A 10-day expert data science diagnostic over a maximum of 3 months: - Technical assessment and concrete use case identification - **25%** funded by BPI France for SMEs (since January 2026) - Cost for the company: approximately 7,500 euros after subsidy A good starting point before applying for larger grants. --- #### France 2030: AI calls for proposals ##### Pionniers de l’Intelligence Artificielle The flagship France 2030 program for AI, managed by BPI France and Inria. Phase | Purpose | Amount | Grant rate Phase 1 | Technical feasibility | 100,000 to 200,000 EUR | 100% grant Phase 2 | Demonstrator | 400,000 to 800,000 EUR | Up to 50% Phase 3 | Industrialization | 3 to 8 million EUR | Up to 50% **Next deadlines:** March 10, 2026 and June 9, 2026. **Priority sectors:** industry, energy, cybersecurity, healthcare, ecological transition. Phase 1 is particularly attractive: 100% grant funding to validate the technical feasibility of your AI project. --- #### Auvergne-Rhone-Alpes regional grants For businesses in the Lyon area, several regional programs complement national funding. ##### R\&D Booster Auvergne-Rhone-Alpes - Project budget: 250,000 to 1 million euros - Grant and/or zero-interest innovation loan - **Artificial intelligence** is a priority area of excellence for the region ##### France 2030 Regionalized in ARA - Over **120 million euros** committed since launch - Parity co-financing State/Region, intervention rate up to **50%** of eligible expenses - Advisory costs: 50% covered, capped at 16,000 euros subsidy per company ##### Atouts Numériques A **100% free** program (funded by the Region + FEDER) for small businesses under 50 employees: - Digital maturity diagnostic - 7 to 14 hours of personalized training - 7 to 14 hours of project follow-up --- #### JEI status: for AI startups The **Jeune Entreprise Innovante** (Young Innovative Company) status offers exemptions from employer social security contributions and family allowance contributions, particularly suited to AI startups whose main costs are engineer salaries. **Conditions:** - SME under 8 years old - At least **20%** of expenses dedicated to R\&D (threshold raised in 2025) - Capital held at least 50% by natural persons - Company created before December 31, 2028 **Benefits:** - Exemption from employer social security and family allowance contributions - On salaries up to 8,203 euros per month - Capped at 240,300 euros per establishment per year - For **7 years** **New in 2026:** the JEII (Young Innovative Impact Company) status is for companies with at least 5% R\&D expenses and a social or environmental mission (ESUS or SSE status). --- #### Practical tips 1. **Distinguish R\&D from integration.** For the CIR, your project must involve genuine technical uncertainty. Document the state of the art, problems encountered, and approaches tested. 1. **Stack programs strategically.** CIR + JEI + BPI France ADI can be combined. The CIR reduces taxes, JEI reduces social charges, ADI provides cash. 1. **Start with a diagnostic.** Diag Data IA or Atouts Numériques are accessible entry points that produce documents useful for subsequent applications. 1. **Apply early.** Earlier submission rounds generally have more budget available. 1. **Detail the budget and ROI.** Justify every expense line and show financial viability beyond the grant period. --- #### Do you already have an AI project in mind? At [Flowful AI](/about/), we build AI solutions for French businesses. Many of [our projects](/projects/) qualify for the programs above, and we structure them so grant applications are easier to put together. Based in [Lyon](/agency/lyon/) and [Brussels](/agency/brussels/), we work with SMEs and mid-caps across France. **[Book a discovery call](/contact/)** to scope your AI project. We can point out which funding routes are likely to apply. The application itself is then best handled by you directly, or through a specialised cabinet. ### GEO & AEO: How to Get Cited by ChatGPT, Perplexity, and AI Search Engines Source: https://flowful.ai/blog/geo-get-cited-by-chatgpt-ai-engines/ Language: en Description: A 10-point checklist to get your business cited by ChatGPT, Perplexity, and AI search engines, from robots.txt and llms.txt to FAQ pages and E-E-A-T signals. You’ve spent years optimizing for Google. Title tags, backlinks, keyword density. The whole playbook. But something has shifted. Your potential customers are now asking ChatGPT, Perplexity, Mistral, and Gemini for recommendations instead of scrolling through ten blue links. When someone asks “What’s the best lead generation tool for small businesses?” and your company doesn’t show up in the AI-generated answer, you’ve lost that prospect before they ever saw your website. **Traditional SEO gets you ranked. GEO gets you cited.** And in 2026, being cited by an AI engine is becoming more valuable than ranking on page one. #### What Are GEO and AEO Let’s clarify three terms that often get mixed up: **SEO (Search Engine Optimization)** targets traditional search engines to rank your pages in the classic “ten blue links.” You already know this one. **AEO (Answer Engine Optimization)** targets direct answer systems like Google’s AI Overviews, featured snippets, and voice assistant responses. When Google pulls a direct answer from your site and shows it above all search results, that’s AEO at work. **GEO (Generative Engine Optimization)** goes further. It optimizes your content specifically for large language models like ChatGPT, Claude, Mistral, Perplexity, and Gemini. These models don’t just show your link. Their retrieval systems surface content that is well-structured, authoritative, and relevant, and when they do, they may cite you as a source in their generated answers. The concept was formalized in a [2024 research paper from Princeton and Georgia Tech](https://arxiv.org/abs/2311.09735), which demonstrated that specific optimization strategies can significantly increase a page’s visibility in AI-generated responses. The good news: GEO, AEO, and SEO are complementary. Most GEO improvements also strengthen your traditional SEO. Structured data, clear content architecture, and semantic HTML help both Google and AI engines understand your site. The key difference: **AI engines don’t send you traffic by default. They extract your information and present it directly.** Your only chance of getting attribution (and the click) is if your content is structured, authoritative, and specific enough that the AI deems it worth citing. This is what GEO is about. Not gaming an algorithm, but making your content so well-structured and useful that AI retrieval systems are more likely to surface and reference you. #### Technical Foundation These are the technical signals that determine whether AI engines can even find and parse your content. Start here. Most of these take hours to implement, not weeks, and they’re the highest-ROI items on this list. You can check all of these instantly with our free **[GEO Ready Score](https://geo-ready.flowful.ai)** tool. It runs 10 checks on your site and tells you exactly what’s working and what’s missing. ##### 1. AI Crawler Access (robots.txt) Your robots.txt file controls which bots can crawl your site. Many websites still block AI crawlers without realizing it. Check that you’re explicitly allowing the major AI fetchers used for citations (search and user-triggered fetch): - **OAI-SearchBot** (OpenAI, ChatGPT Search citations) - **Claude-SearchBot** and **Claude-User** (Anthropic, search + user-triggered fetch) - **PerplexityBot** (Perplexity) - **MistralAI-User** (Mistral, user-triggered fetch) - **Google-Extended** (Gemini training/grounding controls, robots.txt product token) If your goal is citations without training, you can allow the fetch/search agents above while still blocking training crawlers like GPTBot and ClaudeBot. If these bots can’t access your pages, they’re far less likely to cite you. Each provider publishes its crawler details. For example, [Google documents Google-Extended](https://developers.google.com/search/docs/crawling-indexing/overview-google-crawlers) alongside its other crawlers. Also make sure your robots.txt references your sitemap URL so crawlers can discover all your pages efficiently. ##### 2. llms.txt This is an emerging standard that provides context about your website specifically for large language models. Think of it as a README for AI. It tells models what your site is about, what your main pages cover, and how your content is organized. Most websites don’t have one yet, so adding it can give you an early-mover edge. The format is simple and lightweight. You can learn more about the standard at [llmstxt.org](https://llmstxt.org/). ##### 3. JSON-LD Structured Data Structured data is how you translate your content into a language AI engines parse natively. Mark up your content with `FAQPage`, `HowTo`, `Article`, `Organization`, and `Product` schemas. AI engines use these markup types to understand what your page is about and extract facts with confidence. This is table stakes. If you’re not doing this yet, it’s the single highest-ROI item on this list. A properly marked-up FAQ page is dramatically more likely to get cited than the same content without schema. JSON-LD is the preferred format: embed it in your page’s `` and keep it synchronized with the visible content. ##### 4. Semantic HTML Use proper HTML tags (`
`, `