AI Readiness Assessment: What It Checks and Whether You Need to Buy One

I went looking for an AI readiness assessment the way a business owner would, by typing the phrase into Google and clicking what came back. Nine results on page one. Seven of them wanted my name, my work email and in one case my phone number before they would show me a single score.

That tells you what the category has become. An AI readiness assessment is not really a product. It is about twenty honest questions, and almost every version on page one is those questions wrapped around a sales call.

So here is my version, ungated. What these assessments genuinely check, what the paid ones cost in the UK with the sources, the questions I would answer before spending anything, and the one situation where writing a cheque for one is the right call. Fair warning: I think most smaller businesses should run this themselves in an afternoon and put the money into building something instead.

What is an AI readiness assessment?

An AI readiness assessment is a structured review of whether your business has the data, the documented processes, the tools, the internal ownership and the governance to get real value from AI. It produces a score across several dimensions and a list of prioritised recommendations. It diagnoses. It builds nothing.

The shape is remarkably consistent wherever you look. Microsoft’s free assessment on Microsoft Learn scores you across seven pillars, including business strategy, data foundations, organisation and culture, infrastructure, and model management, and takes about forty five minutes. Cisco’s AI Readiness Index uses six pillars and weights infrastructure at 25 percent and data at 20 percent. TDWI’s free version asks roughly seventy five questions across five dimensions.

Read those pillar lists next to each other and the family resemblance is obvious. Strategy, data, infrastructure, governance, talent, culture. Every vendor rearranges the same six nouns.

Why do the free readiness tools feel wrong for a smaller business?

Because they were written for a CIO with a data team, and you are not that. When Cisco weights infrastructure at a quarter of your total score, remember that Cisco sells infrastructure. When an assessment asks how you handle model management and MLOps, that question is aimed at an organisation training its own models, which is not a fifteen person firm buying a Claude subscription.

The second problem is what you get at the end. A maturity level. You finish forty five minutes of questions knowing you are Level 2 of 5, Planning, which is a fact about a spreadsheet rather than a decision you can act on. Nobody hands you the thing you actually wanted, which is a name of a process and a first step.

And then there is the jurisdiction. I read every page ranking on the first page for this term in the UK and not one of them mentions the ICO, UK GDPR or a DPIA. The best structured of them cites the EU AI Act, which for most British businesses is the wrong rulebook entirely. You are being scored against an American enterprise checklist and then sold a call.

Is your business AI ready, or is one process ready?

Readiness is a property of a workflow, not of a company. This is the single most useful reframe I can give you, and it is the opposite of what every pillar scoring model implies.

Your customer data might be a mess while your invoicing runs cleanly through Xero with twelve months of clean history sitting behind an API. On the pillar model you score badly on data and stall. On the real model you have one process that is ready today and several that are not, and you go and automate the ready one.

Gartner put a number near this. It predicts organisations will abandon 60 percent of AI projects through 2026 because the data behind them was not ready, and found 63 percent of organisations either lack the right data management practices or are unsure whether they have them. Those projects did not fail on ambition. They failed because nobody checked the specific data feeding the specific job.

The barrier data says the same thing from the other end. In the Department for Science, Innovation and Technology’s 2025 adoption research, the most common reason UK firms gave for not using AI was no identified need, at 71 percent, ahead of limited AI skills at 60 percent. That is not a technology gap. It is a nobody has pointed at a specific job gap, which is exactly what a per process assessment fixes and a maturity score does not.

The readiness questions worth answering yourself

Take your single most annoying recurring task and put it through these. It takes an afternoon and it costs nothing.

Where does the data live? Is your customer list in a CRM such as HubSpot, Pipedrive or Capsule, or is it Outlook contacts plus a spreadsheet on one laptop? Do you know your top twenty customers by revenue without asking your bookkeeper? Is there one source of truth for pricing, or does each salesperson carry a version in their own folder?

Can you get it out again? Xero exports contacts and reports to CSV and has an API on every plan. QuickBooks Online is the same story. HubSpot allows exports on all tiers, though the API depends on your plan. Google Workspace and Microsoft 365 are both fully addressable. If the answer is that everything lives in PDFs, or a legacy desktop package nobody has updated since 2019, that is your first project and it is not an AI project.

Is a human retyping between two systems? That is usually the highest value and lowest risk automation in the building. Find it before you find anything clever.

Are you already paying for AI you have not switched on? Microsoft 365 Copilot, Gemini in Google Workspace and HubSpot’s built in AI features all sit inside subscriptions many owners already hold. Read your own bill before you shop. A readiness assessment that does not open your billing page first is not being thorough.

Who actually holds the keys? If your outsourced IT provider controls admin access to your own systems, budget time for that conversation before you budget for the build.

Answer those honestly and you have done what the £1,000 version of this assessment does. What you cannot get for free is someone interviewing your staff and writing it up for a board. That is the real thing you are buying when you pay, and it is worth being clear about that before you sign.

What does an AI readiness assessment cost in the UK?

Between nothing and about £35,000, and the gap is almost entirely about who does the interviewing and who writes the report.

Notice what the ladder is actually charging for. The questions are the same at every rung. What changes is how many of your staff get interviewed and how thick the document is at the end.

That gives you a clean rule. Work out what the problem costs you a year first. If the task eats around £6,000 of staff time annually, a £15,000 diagnostic is the wrong instrument no matter how good the consultancy is. Price the problem, then price the help.

Do you need a DPIA before you point AI at customer data?

Often, yes, and this is the part the ranking pages skip entirely. Under UK GDPR a DPIA is required where processing is likely to result in high risk, and the ICO explicitly flags innovative technology, large scale processing of special category data, automated decisions with legal or similarly significant effects, and systematic profiling. Feeding a folder of customer records into a new AI tool usually trips at least one of those.

The ICO also publishes a free AI and data protection risk toolkit as a downloadable spreadsheet. It is optional, you will not be penalised for skipping it, and it does not replace the DPIA. It is still the most useful free readiness artefact available to a UK business, and none of the pages currently ranking for this term mention it.

Two practical points while you are here. Check where your vendor processes data and whether your inputs are used for training, in the terms rather than on the marketing page. And if you are switching on Microsoft 365 Copilot, know that it inherits your existing permissions, so any oversharing already sitting in SharePoint becomes oversharing that Copilot will happily surface in a search result. Audit those permissions first.

Worth knowing that the government tried to build the free small business version of all this. The Department for Science, Innovation and Technology consulted on an AI Management Essentials self assessment tool aimed explicitly at smaller firms, then confirmed it would not publish the draft and would develop refined guidance instead. Do not go looking for it.

When is paying for an AI readiness assessment actually worth it?

When you need the document more than the answer. That sounds cynical. It is not, because there are real situations where the document is the point.

You are in a regulated sector and your compliance team needs a written record showing that a competent third party considered less risky alternatives. A funder, an insurer or a board wants sign off before releasing budget. Or you have five candidate processes, genuine disagreement internally about which comes first, and an outside voice will settle it faster than another meeting will.

What does not justify the fee is not knowing where to start. That is a cheap problem. Take whichever recurring task you would hand over first if you could, run the five checks against it, and you have your answer before lunch. If you want a rough estimate of the hours and money involved before you commit any time at all, my free AI Opportunity Scorecard takes about two minutes.

The uncomfortable version of this: a readiness assessment is the easiest thing in the world to sell, because it is low risk for the buyer, low effort for the seller, and nothing has to work at the end of it. Be suspicious of how comfortable the purchase feels.

Where I would put the assessment budget instead

If you handed me £3,000 and asked me to get your business ready for AI, I would not spend a penny of it on an assessment. I would put an afternoon into the five checks, then use everything left over to build the winning process properly and teach whoever owns it how it works. You would finish the month with something running, and with a person in the building who can change it when the process changes.

That is the whole difference I care about. A readiness assessment leaves you holding a score out of five. A first build leaves you holding a working process and the start of a skill, and the skill is what stops you renting help forever. Claude Code has moved that within reach of people who have never written a line of code, which describes most of the owners I sit down with. There is more about why I work this way if you want the background.

So run the checks this week, on one process, honestly. Then, if you would rather not build the first one on your own, book an AI Opportunity Audit: a fixed fee session where we find your highest impact automation, build a working one live in Claude Code, and you leave with a prioritised roadmap you can carry on without me.