AI Readiness Checklist for Business: Data, Process, Risk

The board wants an AI pilot, the team has three tools open and nobody can say which process should go first. Months later the pilot is either postponed or launched on messy data, with nobody sure what success means. This is the usual story for companies of 20 to 300 people. Below is an AI readiness checklist in four blocks: processes, data, people and risks. Each block ends with a yes or no question, so you can see whether to prepare, pilot or tidy up first.

AI readiness checklist: why so many projects stall before the first result

Projects rarely stall because the model is weak. They stall because the starting conditions were never checked: the process is not written down, the data lives in five places, nobody owns the result and the legal questions arrive late. The rules also keep moving. Regulation (EU) 2026/1744, the Digital Omnibus on AI, changed the start of some high-risk obligations that were due on 2 August 2026, so check the dates that apply to your system. A short checklist of readiness for AI turns this worry into four questions you can answer in an afternoon.

Block 1. Processes: what is written down, repeats and has a measurable goal

A readiness checklist for AI projects starts with one process, not a tool. Check the readiness of your AI project before buying any tool. Pick one process and describe it in a page: who starts it, what happens, where it ends and how long it takes today. A good first candidate repeats weekly, follows clear rules and has a number you can measure, such as hours per request or errors per hundred. Our guide on business process automation shows how to choose one. If the process lives only in someone’s head, write it down before any tool arrives. Question: can you describe one repeating process and its baseline in numbers?

Block 2. Data: where it lives, who owns it and how clean it is

Data readiness comes down to three facts per source: where the data is stored, who owns it and how reliable it is, and supplier invoices are a good test case, as in invoice processing automation. Check duplicates, missing fields and old records, and judge data quality honestly. Note who may see personal data. AI data readiness is not perfection, only knowing which source is good enough for the pilot. Our AI development services start with the same inventory, because a model is limited by what it can reach. Question: do you know the owner and the quality of the data this process needs?

Block	Evidence you need	Red flag
Processes	One page with a baseline	Rules differ by person
Data	Source, owner, quality note	Copies in many files
People	Named pilot owner	Everyone and no one
Risks	Access rules, review step	No one checks answers

Block 3. People: who owns the pilot and how the team will work with it

An AI pilot without an owner drifts, and no checklist can fix that for you. Whoever runs the AI readiness checklist must name the owner first. Name one person who is accountable for the outcome, one who knows the process and one who handles access and security. Then ask how the team will use the result. Who reviews the output, who handles exceptions and who explains it to colleagues? Training matters too. And Article 4 of the EU AI Act asks providers and deployers to take measures for a sufficient level of AI literacy among staff. Question: is there a named owner, and a plan for reviewing the output?

Block 4. Risks and regulation: GDPR, the EU AI Act, access and checking answers

List what could go wrong: personal data sent to a tool without a contract, answers that sound right and are wrong, as in a badly set up AI chatbot for customer service, access that is too wide. Under the GDPR, check the lawful basis and the processor contract for any vendor that sees personal data. Under the AI Act, find out which role you play and whether your use case is limited or high risk. Keep a human review step for anything that touches customers, prices or contracts, and write the governance down. Every readiness checklist should end here with a check. Question: do you know your data limits, your vendors and who checks the answers?

Where risk hides

In tools bought without review.

What to log

Vendor, data, owner.

Reading the result of the readiness check for AI: prepare, pilot or tidy up first

Count the yes answers. Use it as an AI implementation checklist.

1. Four yes answers: you can pilot now.
2. Two or three: prepare for a month and fix the gaps in the table.
3. Fewer than two: tidy up first, describe the process, find the data owner and name the pilot lead.

Our AI consulting services run this AI readiness assessment with you, the same AI consulting for small business we offer to larger firms, and turn the answers into a plan. Before AI touches customers, a short checklist tells you whether the readiness is real. Readiness is a habit; a checklist for AI only records it.

Choosing the first pilot and what to measure

Choose a pilot that is small, repeatable and easy to check, for example sorting incoming enquiries or drafting answers to common questions. Agree the baseline first, then one or two numbers: time saved, error rate or response speed. Run it for a month. Compare. The NIST AI Risk Management Framework offers a neutral vocabulary for risks if you need one. AI workflow automation can then take over once the pilot proves itself, and how to prepare your business for AI becomes a routine. An AI strategy for business then grows from one proven process, and AI adoption follows.