In this article7
An AI readiness assessment checks whether your business is prepared to use AI in a way that actually pays off: is your data in order, do you know how your processes really run, is someone driving it, and are there rules for safe use? It comes down to five questions: where is your data, how good is it, who knows the process, who owns the outcome, and what governance applies. None of these has to be perfect to start, but you do need to know where you stand.
The short version:
- Lack of experience is the biggest barrier to AI adoption: 71.6% of Dutch micro businesses that considered AI but didn't use it cited it (CBS, 2025).
- Readiness is not about technology. It's about data, process knowledge, ownership and governance.
- You don't need high scores everywhere. One well-prepared process is enough to start.
- The output of a readiness assessment is a concrete first use case, not a report for the drawer.
Why AI readiness matters more than the tool
Most money lost on AI doesn't go to bad models, but to projects that land on an unprepared organization. MIT's 2025 report, which found 95% of organizations see no measurable returns from GenAI pilots, named poor integration into the organization as the main cause. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027, partly due to unclear business value and inadequate risk controls.
Governance in particular lags behind ambition. Deloitte's State of AI in the Enterprise 2026, a survey of 3,235 leaders in 24 countries, found that nearly three quarters plan to deploy agentic AI within two years, but only 21% have a mature governance model for agents. Eurostat data shows that among EU companies with 10+ employees that don't yet use AI, only 14% have even considered it.
A readiness assessment turns that uncertainty into concrete questions. For what happens to projects that skip this step, see why AI projects fail.
The AI readiness assessment checklist: five parts
1. Data map: where does your information live?
List the systems that hold your business information. For a typical small or mid-sized company that might be:
- Accounting and ERP: QuickBooks, Xero, NetSuite, Exact or similar
- CRM: HubSpot, Salesforce or a custom system
- Documents: SharePoint, Google Drive, a network share
- Communication: Outlook or Gmail, messaging apps, a helpdesk
- Spreadsheets that are really databases in disguise
For each system, note what data it holds, who manages it and whether there's an API to connect to it. That last question often decides what's feasible.
2. Data quality: can you trust it?
An AI system takes your data seriously, even when it's wrong. Spot-check:
- How many duplicate customers or contacts are in your CRM?
- Are required fields actually filled in?
- Are work instructions, price lists and terms current, and is there only one version?
- Does critical knowledge live mostly in people's heads and inboxes instead of in systems?
Perfect data doesn't exist. But know which data your first use case needs, and clean up exactly that part.
3. Process knowledge: do you know how the work really flows?
The process on paper is rarely the process in practice. Before automating, someone needs to be able to explain:
- The steps, from start to finish
- Which exceptions occur and how often
- How long each run takes and how often it happens per week
- What goes wrong and what an error costs
Those numbers are your baseline. Without one, you can't show later what AI delivered. The math is covered in how to calculate AI ROI.
4. Champion: who drives it?
Every AI initiative needs a business owner. Not an IT person doing it on the side, but someone who knows the process, decides on exceptions and brings the team along. That person needs time: budget a few hours per week during build and testing.
Also ask how your team feels about it. Are there concerns about jobs? Is there experience with AI tools? Adoption starts with the people who do the work today, because they know the exceptions.
5. Governance: what rules are in place?
For a small or mid-sized business, governance doesn't have to be complicated, but a few things must be settled:
- An AI tool policy. Which tools may staff use, and with what data? Regulators such as the Dutch data protection authority have warned that entering personal data into a chatbot against company rules can be a reportable data breach.
- Business accounts. Business tiers of ChatGPT, Claude and Gemini don't train on your data by default. Consumer versions can. See ChatGPT and GDPR.
- Data processing agreements with your AI vendors.
- EU AI Act, if you operate in the EU. Prohibited practices and the AI literacy requirement apply since February 2, 2025, and transparency rules for chatbots and similar systems since August 2, 2026.
Score yourself: the readiness matrix
Give yourself a score per part. The goal isn't a high total, but knowing where your first step is.
| Part | Not yet (1) | Partly (2) | In place (3) |
|---|---|---|---|
| Data map | No overview of systems | Overview exists, integration options unknown | Systems and APIs mapped |
| Data quality | Many duplicates, outdated documents | Quality varies by system | Data for the first use case is reliable |
| Process knowledge | Process lives in people's heads | Steps known, no numbers | Steps, exceptions and baseline known |
| Champion | Nobody assigned | Someone willing, no time | Owner with time and mandate |
| Governance | No rules on AI use | Informal agreements | Policy, business accounts, processing agreements |
5 to 8 points: start with the basics. Pick one process and map its data and numbers. 9 to 12 points: you can run a scoped pilot while fixing weak spots in parallel. 13 to 15 points: you're ready to build. Pick the use case with the highest baseline cost.
From readiness to a first project
An assessment is only useful if it leads to a decision. Take the result and pick one process that:
- Runs often (dozens or hundreds of times per week)
- Demonstrably costs time today
- Has a clear correct outcome
- Relies on data that is already reasonably clean
- Has an owner
That's your first use case. The full path from there to production is in how to implement AI in business.
Frequently asked questions
What is an AI readiness assessment?
An AI readiness assessment is a structured check of how prepared your business is to use AI. It looks at your data and systems, the quality of that data, how well your processes are understood, who owns the outcome and what rules exist for safe use. The output is a concrete first use case and a list of what still needs fixing.
How do I know if my company is ready for AI?
If you know where your data lives, the data for one process is reasonably reliable, you can describe that process with numbers, there is an owner and you have basic rules for AI use, you're ready to start. You don't need perfect scores everywhere. One well-prepared process is enough for a first pilot.
Does my data need to be perfect before starting with AI?
No. Perfect data doesn't exist, and waiting until everything is clean means never starting. The data your first use case depends on does need to be reliable enough. Clean up exactly that part and expand as you go.
How much does an AI readiness assessment cost?
It varies by provider and depth. Airflows offers a free AI scan as a first exploration. If you want to go further, a Discovery (EUR 3,500) works out the use case, data, integrations and baseline as the foundation for a build.
Next steps
Rather not score yourself? Take our free AI scan for an outside view of where your opportunities are and what to fix first. For how we take it from there, from Discovery to a working MVP in four to eight weeks, see our services.
Sources
- CBS: Use of AI technology by Dutch micro businesses, March 16, 2026
- Eurostat Statistics Explained: Use of artificial intelligence in enterprises, December 11, 2025
- Deloitte: State of AI in the Enterprise 2026, January 21, 2026
- Fortune: MIT report: 95% of generative AI pilots at companies are failing, August 18, 2025
- Gartner: Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, June 25, 2025
- Autoriteit Persoonsgegevens: Use of AI chatbots can lead to data breaches, August 6, 2024
- AI Act Service Desk: Timeline for the Implementation of the EU AI Act, September 25, 2026
- OpenAI: Enterprise privacy and Claude: Plans & Pricing, accessed September 26, 2026
Builds AI agents, automations and custom software for businesses at Airflows.
