AI readiness checklist for companies: six checks

What does AI readiness actually mean?

Readiness is easier to judge with a list. Here is the checklist we would use.

Companies often ask whether they are ready for AI as if it were a yes or no. It is not. Readiness is a set of practical conditions, and most firms meet some and miss others. Missing one does not stop you. It tells you what to fix first. Treat the checklist as a map, not a test you pass or fail.

The value of a checklist is honesty. It makes you look at the data nobody owns, the process that lives in one person's head and the risk nobody has discussed. Better to find these now than halfway through a build.

Is your data ready for AI?

Where does it live?

Data comes first, because AI can only work with what it can reach. Check three things about it.

• Put the checklist to work: score each area, rank the gaps by their effect on the first use case, and book time with the sponsor to agree what to fix first
• Where it lives: list the systems, spreadsheets, inboxes and shared drives that hold the information for your chosen task, and who owns each one
• Whether it is good enough: look at a sample for gaps, duplicates and out-of-date records, and note what would need tidying first

Is it good enough?

Processes are the second area. If the work is done differently by each person, or lives in someone's head, there is nothing stable to build around. Write down the steps, the hand-offs and the exceptions. A simple map of who does what and where it waits is enough to start.

Are your processes clear?

People and skills come third. Look at who will use the solution, who will review it and who will look after it. Typical questions include the following.

• Users: do the people who will work with it have time to learn, and have they been asked what would actually help them in their week
• Reviewers: who will check the output, and do they have the knowledge and the authority to reject something that is wrong or unclear
• Owner: one named person responsible for results and for raising problems
• Champions: respected colleagues who will try it first and show others how it fits into daily work
• Training: a plan for teaching the team what to trust, what to check and where to ask for help

Fear is the hidden blocker. Say plainly what will change and what will not, and involve people early. Resistance usually shrinks when staff see AI take the tedious work away.

Are your people ready for AI?

Security is the fourth area. Before any tool touches company information, know what data it will see, where it is processed, who can read the results and how long records are kept. Check each provider's terms, and set access so the AI sees only what the task needs.

Add the basics of agent safety: restricted tools, approval for actions that matter, a log of what was done and a way to switch it off. These are easier to build in at the start than to add after something goes wrong.

Is security in place?

The fifth area is a sponsor. AI projects touch several teams, so someone senior must back them. The sponsor sets the goal, protects time, settles disputes between teams and decides when to expand or stop. Without one, a pilot can run well and still go nowhere because no one has the authority to move it into daily work. Choose someone who cares about the outcome, not only the budget, and agree how often they will review progress.

A sponsor does not need to understand the technology. They need to understand the problem and be willing to back the change.

Do you have a sponsor?

The sixth area is the first use case, and it matters most. Choose a task that is frequent, slow and low risk, where a mistake is easy to catch and the data is reachable. Avoid rare tasks, vague goals and anything where an error is costly. A small, visible win builds trust for everything that follows.

What is the first use case?

Honest gaps are normal. Typical ones are data spread across too many places, processes nobody has written down, no one with time to own the result and security rules that do not mention AI yet. Rank the gaps by how much they would hurt the first use case, and fix those first. Leave the others for later. A company that knows its weak spots and works around them is more ready than one that assumes it has none. Rules on data and AI differ by country and sector, so involve your own advisers where the stakes are high.

A simple score helps. Rate each of the six areas from one to three, with one meaning not in place and three meaning solid. Anything at one needs attention before you start. Do not chase a perfect total.

How do you score it?

• Pick one workflow that is frequent and has clear steps
• Measure how long it takes and how often it goes wrong today, as your baseline
• Test on real examples taken from your own work
• Name the person who owns the outcome

What is the Embedded Week?

• We sit with your team for a week
• We watch how the work, data and approvals really flow
• We map the gaps: the copying, the hand-offs, the waiting for answers
• We build a custom solution for your company, then train your team to run and improve it

Book a meeting to start your readiness review

No two companies are ready in the same way. A law firm, a logistics business and a retailer each have different data, different risks and different habits.

That is why we begin with the Embedded Week, not a questionnaire. We see your tools, your data and your habits for ourselves, build a solution made for your company and train your people to run it. To discuss it, book a meeting and we will tell you honestly how ready you are.