What is the difference between an AI readiness assessment and an AI readiness audit?

The words are often used interchangeably, but there is a useful distinction. An assessment is self-reported: you answer questions about your own business and the result reflects what you believe to be true. A strict audit involves someone inspecting the actual evidence, opening the CRM, querying the customer table, reading the process documentation, checking system access. An audit is more rigorous and considerably more expensive. An assessment is faster and, answered honestly, still surfaces the biggest readiness gaps, because most of those gaps sit in the difference between what leadership assumes and what is actually the case. A well-written assessment question exposes that difference the moment you try to answer it.

How is an AI readiness score calculated, and what is a good score?

In a well-built assessment, each question contributes points to one of six pillars, and each pillar is scored and banded separately rather than merged into a single percentage. That matters because readiness does not average. A business with excellent data, modern technology and a clear strategy but no accountability for process is not 80% ready, it is blocked on process. The overall total is useful mainly as a baseline to restate later. There is no universal good score. The number that matters is the lowest band, because that is the constraint that determines what happens to your next pilot regardless of how strong the other five areas look.

Is a free AI readiness assessment worth taking, or do you need a paid one?

A free assessment is worth taking if it is deep enough to separate one problem from another. Six generic questions cannot distinguish a data problem from a governance problem, so the resulting score tells you nothing actionable. Thirty specific questions across six pillars can. What a free tier will not give you is an action plan, an inspection of your systems, or a review of the tools you already pay for. It gives you the shape of the problem and where the constraint sits. That is usually enough to decide what to fix first. Pay only when you want the sequenced plan rather than the diagnosis.

14 September 2026

What is an AI readiness assessment, and what should it cover?

A plain definition of an AI readiness assessment: the six areas it should cover, how scoring works, and what a self-reported answer can and cannot prove.

By Carl Chessum

An AI readiness assessment is a structured set of questions that establishes whether the conditions AI needs already exist in your business, before you spend money finding out the hard way. It covers the same ground every time: your data, your processes, your people, your technology, your strategy and your governance. It produces a score, usually banded by area, so you can see which of those six is the one that will sink you. It is a diagnosis, not a treatment.

That is the whole definition. Everything else is detail about depth. A good assessment asks enough specific, uncomfortable questions that you cannot answer them all optimistically without noticing you are doing it. A bad one asks six vague questions about “AI maturity”, scores you 72%, and tells you that you are an “emerging adopter”. Both take a few minutes. Only one of them changes what you do on Monday.

What an assessment is, and how it differs from an audit

People use the two words interchangeably and mostly that is fine. But there is a real distinction worth holding.

An assessment is self-reported. You answer questions about your own business, and the output reflects what you believe to be true. An audit, in the strict sense, involves someone inspecting the actual artefacts: opening the CRM, running a query against the customer table, reading the process documentation, checking who has admin rights on which system.

The self-reported version is faster, cheaper and, handled honestly, surprisingly accurate about the things that matter most. The reason is that the biggest readiness problems are not hidden in the data. They are hidden in the gap between what leadership assumes and what is actually the case, and a well-built question surfaces that gap the moment you try to answer it. Ask a CEO how complete their customer data is and they will say “fairly good”. Ask them how confident they are in a clean, accurate list of their best 500 customers if they needed it right now, and the answer changes, because the second question has a deliverable attached to it.

That is the craft in assessment design. Not the scoring model. The questions.

What it should cover: six areas, not three

Most free scoring tools online cover three things, and they are almost always technology, strategy and “culture”. That is the vendor’s view of readiness, because technology is what they sell, strategy is what justifies the engagement, and culture is what they blame when it does not work.

A readiness assessment worth the time covers six.

Data. Whether the information exists, whether it is accurate, whether systems can share it without someone exporting a spreadsheet, and whether anybody is actually accountable for its quality. Dirty data is the cheapest AI mistake to fix and the most expensive to ignore, because every downstream failure gets blamed on the model instead.

Process. Whether your core processes are documented well enough that a new starter could follow them unaided. Whether three people asked to do the same task would do it the same way. Whether anyone reviews processes rather than just running them. You cannot automate a process that only exists in someone’s head, and you certainly cannot measure the improvement.

People. Your team’s actual attitude to AI, their actual capability if you asked them to use a tool tomorrow, and whether you as a leader are modelling adoption or delegating it. Also whether you have made any formal investment in helping them understand it.

Technology. What you have, whether it connects, and whether it can support anything beyond a demo.

Strategy. Whether AI is attached to a business outcome with a number on it, or whether it exists to satisfy a board question. Writer’s 2026 research found that 75% of executives admit their company’s AI strategy is “more for show” rather than real internal guidance, and 48% call AI adoption a massive disappointment, up from 34% the year before. Those two numbers are the same finding said twice.

Governance. Who decides, who approves, who is accountable when something goes wrong, and what your people are already using without telling you. Deloitte’s 2026 Global CSO Survey found 65% of chief strategy officers do not own the top strategic decisions in their own function. If the strategy function does not own strategy, the question of who owns AI outcomes in your company is not rhetorical.

Drop any one of the six and the assessment becomes a story about the five that are fine.

Why scoring matters more than the score

A single overall percentage is almost useless. Readiness does not average.

If your data is excellent, your technology is modern, your strategy is clear, and nobody in the business is accountable for whether processes are followed, you are not 80% ready. You are blocked on process and everything else is irrelevant until that changes. An averaged score hides exactly the thing you needed the assessment to find.

So the scoring should be banded per area. Six bands tell you where the constraint sits. One number tells you how you feel about yourself.

The second thing scoring gives you is a baseline you can restate later. This matters more than it sounds. MIT’s research found 95% of AI pilots deliver zero measurable P&L impact, and the most common structural reason is not that the model underperformed. It is that nobody wrote down the number the pilot was meant to beat, so when the pilot “went well”, there was nothing to compare it to and nothing to defend at budget time. A scored assessment taken before you start is one of the few baselines that costs nothing to establish.

What a self-reported assessment cannot tell you

Be clear-eyed about this, because vendors rarely are.

A self-reported assessment does not inspect your systems. It does not query your database, read your code, check your access logs or verify a single claim you make. Every answer is yours.

It does not review the tools you have already paid for, or tell you whether anyone still uses the licences you renewed in January.

It does not tell you whether your people are already using AI you never approved. It can only tell you whether you think they are. PagerDuty’s 2026 Shadow AI survey of 1,250 office professionals at companies with $500m or more in revenue found two-thirds had used unauthorised AI tools at work, while only 18% of organisations report formal AI security policies. If you are confident that does not describe you, the question worth asking is how you know, and whether shadow AI is a real risk at 500 people is a reasonable place to start.

And a free assessment does not produce a plan. Diagnosis and treatment are different products, and any tool that claims to give you both in the time it takes to make coffee is giving you neither.

What it does give you, reliably, is the shape of the problem. That is worth having, because most businesses that have spent money on AI still cannot describe the shape of their problem. PwC’s 2026 CEO Survey found 56% of CEOs report neither increased revenue nor decreased costs from AI in the last 12 months, with only 12% reporting both. Those 56% are not short of AI. They are short of a diagnosis.

How honest do you have to be?

Completely, or do not bother.

This is the one failure mode of a self-reported instrument, and it is entirely in your control. The temptation is to answer as the business would look in a board pack. The value comes from answering as it looks on a Tuesday afternoon when someone is off sick.

A useful discipline: for any question where you are tempted to pick the second-best option, ask who in the business would contradict you, and whether they would be right. If the answer is “our ops manager, and yes”, pick the honest option. Nobody sees the result but you.

What the free assessment here actually is

Specifics, so you know what you are getting before you click.

It is free. It is 30 questions, five in each of the six pillars, in this order: Data, Process, People, Technology, Strategy, Governance. It takes 7 minutes. It is a form, not a call, and no sales call is booked at the end unless you ask for one.

At the end you get a readiness score out of 120, a band for each of the six pillars, and a short synopsis. That is the whole free tier. No action plan, no roadmap, no inspection of your systems.

The questions are deliberately concrete. The first Data question asks how confident you are in a clean, accurate list of your best 500 customers if you needed it right now. A Process question asks how similar three people’s approaches would be if you asked them all to complete the same task. A People question asks how actively you personally are modelling AI adoption, trying it, talking about it, normalising it, rather than asking whether “the culture is supportive”. Each one has a real answer that you either have or you do not.

That depth is the point, and it is the difference between this and a 60-second scoring widget. Six questions cannot separate a data problem from a governance problem. Thirty can.

If you want the plan as well as the diagnosis, there are paid tiers. The Full Report is £97 and includes action plans. The Deep Dive is £497 and is a 30/60/90-day plan for fixing the gaps before you spend on AI. Both sit on the same diagnosis. You do not need either to get value from the free one, and plenty of people take the free assessment, read the band that came back lowest, and go and fix that themselves.

When to run one

Before you fund the next pilot, obviously. But also after a pilot that failed, because the post-mortem question is rarely “was the model any good” and usually “which of the six conditions was missing”.

Also run one when the board asks for an AI strategy and you need something more substantial than a slide. A banded score across six areas with a named constraint is a better answer than a roadmap nobody has costed. S&P Global Market Intelligence found the share of firms scrapping most of their AI initiatives rose to 42% in 2025 from 17% a year earlier, with the average organisation scrapping 46% of its proofs of concept before production. Those write-offs were funded by roadmaps. They were not prevented by diagnoses, because nobody ran one.

If you have read what an AI readiness audit actually measures and want to stop reading about it, the free AI readiness assessment is 30 questions and 7 minutes, and the worst outcome is that you find out you were right about yourself.