An AI readiness assessment answers one question: where will AI pay off in this business, and what stands in the way? It looks at five things in turn: business goals, data, systems, people and risk. The output is a ranked list of use cases with effort, risk and return for each, plus the gaps you need to close before building.
It is not a maturity score or a vendor pitch. A good assessment can conclude that a company should not build anything yet, and say exactly why.
When you need one
You need an assessment when the pressure to "do something with AI" is ahead of the plan. The common signs:
- Leadership has asked for an AI strategy and nobody owns it yet.
- Several teams have proposed AI projects and they compete for the same budget and engineers.
- A pilot impressed people in a demo but has not reached real users.
- You are about to sign with a vendor and cannot say what success looks like.
- Your data sits in systems that nobody has checked for quality, ownership or permission to use with an external model.
If you already have one validated use case, clean data and a team ready to build, skip the assessment and go straight to scoping the build. The assessment exists to reduce uncertainty. When there is little uncertainty left, it adds little.
What an AI readiness assessment covers
Each area below produces a finding, not just a rating. A finding says what is true today, why it matters for AI, and what it would take to fix.
1. Business goals and use cases
The assessment starts from the outcomes the business is measured on this year: cost to serve, sales cycle length, response times, compliance incidents, retention. Every candidate use case has to map to one of them.
The candidates come from interviews with leadership and with the people who do the work. Front-line teams know where hours are lost to searching, copying, re-keying and answering the same question twice. Leadership knows which of those problems is worth money.
Each use case is then scored on value, feasibility, risk and time to value. The scoring is simple on purpose. Its job is to make trade-offs visible so leadership can argue about the right things.
2. Data readiness
This is where most assessments find the real constraints. For each shortlisted use case, the review checks:
- Where the data lives and who owns it.
- Whether it is complete, current and consistent enough for the task.
- Whether there is enough labelled history to measure quality, such as resolved tickets or approved documents.
- Whether you are allowed to send it to an external model provider, or need private deployment or local models.
- How it will stay current once the system is live.
A use case with high value and poor data is not rejected. It is moved later in the plan, behind the data work it depends on.
3. Systems and integration
An AI feature is only as useful as the systems it can read from and write to. The review maps the CRM, ticketing, document stores, telephony, databases and internal tools each use case would touch, and checks for usable APIs, access controls and audit logs.
Integration effort is often larger than the AI work itself. Finding that out in week two of an assessment is far cheaper than finding it in month three of a build.
4. People and skills
Someone has to own each system after launch: watch quality, handle incidents, decide on changes. The assessment looks at who that would be, what skills the team already has in data, engineering and product, and where you would need to hire, train or bring in outside help.
It also looks at the people who will use the system. A support agent who does not trust an AI draft will rewrite it every time, and the time saving disappears.
5. Risk, security and compliance
For each use case the review asks what happens when the AI is wrong, who is affected, and how errors would be caught. It also checks the rules that apply to your data, such as GDPR or HIPAA, and what your security team will require before anything goes live.
The output is a set of guardrails agreed in advance: which data can be used, how outputs are checked, who signs off, and how incidents are reported. Leaving this to the end is one of the most common reasons good pilots never ship.
What you get at the end
A useful assessment ends with documents a leadership team can act on in the next meeting. In my engagements that means:
- A ranked shortlist of high-value use cases, each with effort, risk and expected return.
- A data, systems and skills readiness review, with the gaps that block each use case.
- A written report and an executive briefing.
The ranking is the important part. It turns a long list of ideas into a decision about what to fund first, what to fix first, and what to drop.
How long it takes
A focused readiness assessment takes two to three weeks. The first week is interviews and access to systems and sample data. The second is analysis and scoring. The third, when needed, is review with leadership and the final briefing.
Longer is rarely better. An assessment that runs for months tends to drift into a strategy document that nobody acts on. If the shortlist is clear and leadership agrees with it, the next step is either a sequenced roadmap across several use cases or a scoped build of the top one.
An AI readiness checklist you can run yourself
Before bringing anyone in, you can get a rough answer with these questions. If you can answer most of them with confidence, you may not need a formal assessment.
- Can you name the two or three business outcomes AI should move this year?
- Do you have a shortlist of use cases tied to those outcomes, with a rough value for each?
- For the top use case, do you know where the data lives, who owns it and whether you may use it with an external model?
- Do the systems involved have APIs and access controls you can build on?
- Is there a named owner for the system after launch?
- Do you know how you would measure whether its answers are good enough?
- Has your security or compliance team agreed what data and deployment options are acceptable?
- Is there budget for a staged plan rather than a single large bet?
Every "no" is a finding. Three or more usually means an assessment will save money, because those are the gaps that turn a promising pilot into a stalled one.
Common mistakes
- Treating the assessment as a maturity score instead of a decision about what to build.
- Interviewing only leadership and missing where the work actually happens.
- Skipping the data check because the demo worked on sample data.
- Picking a model or vendor before the use cases are ranked.
- Producing a long list of opportunities with no order and no owner.
Where to go from here
If you want to start on your own, the checklist above and the five-step approach in my guide on how to build an AI strategy for a mid-size company cover most of the ground. If you already have a pilot that has stalled, the article on why AI pilots stall and how to get them into production is the better starting point.
If you want an outside view, my AI Readiness and Strategy Assessment takes two to three weeks and ends with a clear, ranked view of where AI is worth investing in your business.