AI Readiness Assessment: Is Your Business Ready for AI?

By Grocito15 October 2026

An AI readiness assessment is a short, honest check of whether your business can get value from AI today. You are ready when you have a clear problem worth solving, data you can access and trust, a documented process, people who will use and own the result, basic security and privacy controls, and a budget that allows for iteration. Gaps are normal and fixable.

Key takeaways

Readiness is about the problem and the foundations, not about owning the newest tool.
Six dimensions matter most: problem, data, process, people, governance and budget for iteration.
A weak area does not mean "never"; it tells you what to fix first or which small pilot to choose.
Pick a first use case that is narrow, frequent, measurable and low risk.
Write success criteria before the pilot starts, so you can judge it fairly.
If you are not ready, tidying data and documenting a process is real progress, and it pays off even without AI.

What does AI readiness actually mean?

Readiness means that if you switched on a sensible AI tool next month, it would have something useful to work with and someone accountable for the outcome. Many failed AI projects did not fail because the technology was weak. They failed because nobody owned the result, the data was scattered, or the process being automated was never written down.

Think of six ingredients:

A clear problem. A specific task that costs time, money or quality today.
Data. Information the AI can reach, that is accurate enough, and that you are allowed to use.
Process. Steps people can describe, so you know what to improve.
People. Users who are willing, plus a named owner.
Governance. Rules on privacy, security, review and acceptable use.
Budget for iteration. Money and time to adjust after the first version, not just to launch it.

How do you run a self-assessment?

Answer each question below with yes, partly or no. Do not chase a score. The point is to see where the "no" and "partly" answers cluster.

Problem and value

Can you describe the problem in one sentence without mentioning AI?
Do you know roughly how often it happens and who is affected?
Would you know what "better" looks like, for example faster replies or fewer errors?

Data readiness

Do you know where the relevant data lives (systems, spreadsheets, inboxes)?
Is the data reasonably accurate and consistent, or full of duplicates and gaps?
Is there a named owner for it, and do you have the right to use it for this purpose?
Can it be exported or reached through an integration, rather than being locked in someone's head or on paper?

Poor data is the most common blocker. Our guide to data quality basics explains how to find and fix the worst problems before you build anything on top.

Process readiness

Are the steps of the task written down, or could someone map them in a short workshop?
Do you know where the exceptions and hand-offs happen?
Is the process stable, or does it change every week?

AI applied to an undefined process mostly produces faster confusion. If the answers are "no", start with business process mapping.

Skills and change management

Is there a person who will own the pilot and the outcome?
Do the people who do the work today support the idea, or have they not been asked?
Is someone available to review outputs and give feedback in the first weeks?

Security and privacy readiness

Do you know which data is sensitive (customer details, health, finance, contracts)?
Do staff know what they may and may not paste into AI tools?
Do you have a written policy? If not, our AI usage policy template is a quick start.
Can you control who has access, and review what the system does?

How do you choose a first use case?

Good first use cases share a pattern. They are frequent (so you see results quickly), narrow (so they are easy to test), measurable, and forgiving (a human can check the output before it matters). Examples include drafting replies for a support team to approve, sorting incoming enquiries, summarising meeting notes, or extracting fields from standard documents.

Weak first choices are those that are rare, high stakes or vague, such as "use AI to transform sales". Score ideas on four questions: How often does this happen? How easy is the data to reach? How bad is a mistake? Can we measure improvement? Start where the answers are "often, easy, mild, yes".

For ideas on where to begin, see AI workflow automation: where to start.

Should you build, buy or partner?

Buy when a ready-made tool fits the job and your data can be used within its terms. It is quickest to try.
Partner when you need to connect AI to your own systems and documents, or want help designing the pilot. A good partner explains trade-offs and hands over what they build.
Build when the capability is central to your product or your data cannot leave your environment, and you have the engineering team to maintain it.

Many businesses start by buying or partnering and build later, once they know what they actually need. Our build vs buy decision framework applies here too.

How do you design a pilot with real success criteria?

A pilot is a small, time-boxed test, not a half-launched product. Decide these things first:

Scope: one task, one team, one data source.
Baseline: how long it takes or how often errors happen today, measured before you start.
Success criteria: the change you would call worthwhile, agreed in advance, such as shorter handling time with no drop in quality.
Human review: who checks outputs, and what happens when the AI is wrong.
Stop conditions: what would make you pause, such as a privacy concern or poor accuracy.
Decision date: when you will choose to expand, adjust or stop.

What are the red flags?

AI for its own sake. No problem statement, just a wish to "do something with AI".
No owner. Everyone is interested and nobody is accountable.
No data, or data nobody trusts. The tool has nothing reliable to learn from or look up.
Skipping the people. Staff first hear about it at launch.
No review step for outputs that reach customers.
Unlimited expectations. Hoping for a replacement of a whole team rather than help with a task.

What does a phased roadmap look like?

Prepare. Choose the use case, name the owner, tidy the key data, write the usage policy.
Pilot. Run the small test with measured criteria and human review.
Refine. Fix what the pilot exposed: prompts, data, process steps, training.
Extend. Add the next team or the next task, reusing what worked.
Govern. Review accuracy, cost, security and use on a regular rhythm.

What if you are not ready yet?

That is a useful result. Pick the two weakest dimensions and work on them. Clean one dataset. Map one process. Write a one-page AI policy. Nominate an owner. These steps improve operations regardless of AI, and they make a later pilot far more likely to succeed. If you want an outside view, our AI and ML services include discovery work to help you decide what to try first.

A worked example

Imagine a small training company whose support inbox is crowded with repeated questions about schedules and certificates. It runs the self-check and finds: a clear problem (yes), data in a shared drive and a booking system (partly, with duplicates), a rough process (partly), no owner (no), no AI policy (no). Instead of buying a tool at once, it names the operations manager as owner, cleans the schedule data, writes a short policy and maps how enquiries are handled. Then it pilots AI-drafted replies that a person approves, measuring handling time and correction rate against the baseline. Because the foundations were in place, the pilot gave a clear answer.

FAQ

How do I know if my business is ready for AI?

You are ready when you can name a specific problem, reach reasonably clean data, describe the process, assign an owner and set basic rules for privacy and review. If some of those are missing, fix them first or choose a small pilot that works around them.

What is the first thing to do before adopting AI?

Define one narrow, measurable problem and record how it performs today. Without a baseline you cannot tell whether AI helped. After that, check that the data is accessible and that you have a simple usage policy.

Do small businesses need a data scientist to start with AI?

Not usually. Many useful first projects use existing tools or a partner to connect AI to your documents and systems. You do need someone inside the business who owns the outcome and reviews the results.

How long should an AI pilot run?

Long enough to see enough real cases to judge it fairly, but short enough to stay a test. Agree the length and the decision date at the start, and check the results against your success criteria.

Next steps

Run the self-assessment with two or three colleagues this week, mark each answer yes, partly or no, and list the three biggest gaps. Choose one narrow use case and write its success criteria. If you would like a second opinion or help planning a pilot, you can contact us.