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.
Budget realistically.
A focused AI automation (email routing, document extraction, a customer-facing 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. 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.
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 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 and we’ll walk through your specific situation, map the opportunities, 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 or check out Custom Tools vs Enterprise SaaS.