Key points
- An assessment ends in a decision and a plan, not a slide deck of possibilities or a vendor demo.
- It looks at six areas: strategy and goals, data, technology, team and skills, governance and risk, and delivery and measurement.
- You should come away with scored use cases, a data readiness view, a build-vs-buy call and a prioritized 90-day roadmap.
- If you have one obvious, low-risk use case and someone to own it, a small pilot will teach you more than an assessment.
- A free self-assessment takes a few minutes and tells you which areas to look at first.
What is an AI readiness assessment?
An AI readiness assessment answers a narrow question: given where your company is today, which AI use cases are worth doing first, and what has to be true before they can work in production?
In practice, the use cases worth assessing for most growing companies are LLM applications and agents: assistants for internal teams, search and question-answering over your own documents, and workflows that draft, summarize or route work across your systems. The assessment checks whether your data, platform, people and guardrails can support them.
It is short and bounded. A good one takes a couple of weeks, not a quarter, because its job is to get you to a decision quickly.
What an AI readiness assessment is not
It is not a vendor demo. A demo shows what a product can do on clean sample data. An assessment asks what will happen on your data, inside your workflows, with your legal and security constraints.
It is not a strategy deck without a plan. If you finish with a list of "opportunities" and no owners, order, estimate or way to measure success, you have paid for a brainstorm.
It is also not a technology audit for its own sake. The point is not to grade your stack. The point is to find out what stands between a specific use case and real users, and how much work it will take to remove it.
What does an AI readiness assessment examine?
A useful assessment covers six areas. Weakness in any one of them is usually what stalls a pilot, so it pays to look at all six rather than only the technology.
- Strategy and goals: which business outcomes you want AI to improve, who owns each one, and whether there is a sponsor and budget.
- Data: where the data for each use case lives, how easy it is to reach, whether you trust its quality, and who is allowed to see what.
- Technology: whether your platform can integrate new services, and whether you have shipped any AI or LLM feature to real users.
- Team and skills: who would build and run AI features, and how ready the teams who will use them are to change how they work.
- Governance and risk: whether you have an AI usage policy, how sensitive data is handled, and how new use cases get reviewed.
- Delivery and measurement: how you would know a feature works, through business metrics and evaluations of output quality, and how you decide build versus buy.
What should you get out of an AI readiness assessment?
Judge an assessment by what you can do the week after it ends. At minimum, you should have:
If any of these are missing, ask why before you sign. The roadmap and estimate matter most, because they turn the findings into something your leadership can approve or reject. For more on the build-or-buy question, see our guide to building vs buying LLM features.
- Interviews with leadership and the teams closest to the work, so the findings reflect how work actually gets done.
- A use-case inventory, scored by value, feasibility and risk.
- A data readiness review of the systems those use cases depend on.
- A build-vs-buy recommendation for the top use cases.
- A prioritized 90-day roadmap with a build estimate.
- An executive readout your leadership team can act on.
Do you need an AI readiness assessment, or should you just start a pilot?
Not every company needs an assessment. Sometimes the fastest way to learn is to build something small and put it in front of a few users. Use these signs to decide.
You probably need an assessment if your team has run pilots that never reached a product or workflow people use, if the board is asking for an AI plan and you do not have a credible answer, if nobody can say whether your data is ready, if you are unsure whether to build, buy or wait, or if legal and security want guardrails before anything touches customer data.
You should probably just start a pilot if you have one clear use case with an owner, the data it needs is already accessible and not sensitive, the risk of a wrong answer is low, and you can define a metric and a decision date up front. An internal assistant over non-sensitive documentation is a common example. Run it for a few weeks, measure it, and decide.
How to run a quick AI readiness self-assessment
Before you bring anyone in, spend a few minutes scoring yourself. The free AI Readiness Scorecard asks twelve questions, two for each of the six areas above, and places you at one of four levels: Exploring, Building foundations, Ready to pilot, or Ready to scale. You also get a score for each area.
Read the result as a map of where to look, not a verdict. A low score in governance or data is the most common reason a promising pilot stalls, and both are cheaper to fix before you build than after. If you land at Ready to pilot or Ready to scale, you may not need a formal assessment at all.
If you do the exercise with two or three colleagues and compare answers, the disagreements are often more useful than the scores.
Frequently asked questions
How long does an AI readiness assessment take?
Ours takes two weeks: interviews, a data review and use-case scoring with your team, followed by a roadmap and an executive readout. Anything much longer is usually drifting into strategy work without a plan.
Who from our company needs to be involved?
An executive sponsor, the leaders of the teams whose work the use cases would change, and whoever owns your data and platform. Legal or security should join at least one conversation so guardrails are designed in rather than bolted on.
Does the assessment need access to our data?
It needs to understand how your data is stored, governed and accessed, which is usually done through read-only access or walkthroughs with your team. You should sign an NDA before any access is granted.
What happens after the assessment?
You take the 90-day roadmap and start on the first use case. You can deliver it with your own engineers, or bring in outside help to build it with them. Either way, the build is scoped from the roadmap, so you know the estimate before you commit.
What if the assessment shows we are not ready?
That is a useful result. You will know which gaps to close first, usually data access, a usage policy or a way to measure quality, before you spend anything on a build.
Want a senior second opinion on your situation?