You have decided the business needs an outside view on where it stands with AI. An hour of searching later you have a Big Four maturity framework, three boutique consultancies promising practical roadmaps, a free scorecard from a software vendor and a fixed-price diagnostic from a firm you had never heard of yesterday. Each promises a score and a roadmap. None gives you a way to compare what you would actually receive for the money.
This is a buying guide for that moment. The aim is to help you tell the offers apart before you spend anything, whoever you end up hiring.
What are you actually buying with an AI readiness assessment?
Every provider sells the same headline, a diagnostic that scores your business across dimensions such as strategy, data, infrastructure, skills and governance, then hands you a roadmap. The difference lies in the output. One kind of assessment commits you to a prioritised, costed plan tied to your operations. The other gives you a maturity score with a sales deck attached.
Credible frameworks converge on the same ground. Elevated Signal’s 2026 analysis of the market resolves them to six dimensions, strategy, data, infrastructure, talent, governance and culture, and notes that Cisco’s, Microsoft’s and Gartner’s models differ mainly in presentation. The useful question for an owner-managed business is what the assessment does with those dimensions. A genuine diagnostic inspects samples of your operational data, walks through your systems, interviews the people doing the work, and checks your governance against real obligations, including the ICO’s guidance on AI and data protection and the NCSC’s advice on where accountability for AI risk should sit.
The finished article should be a shortlist of two or three first use cases drawn from your actual workflows, each with indicative costs, a named owner and a way to measure the result. I have written separately about what a readiness assessment designed for smaller firms should cover. This post is about judging the people offering to run one.
When is a paid, independent assessment the right call?
A paid, independent assessment earns its fee when you have a specific use case worth shipping in the next six to twelve months and the cost of getting it wrong is material. At that point you need someone to inspect your actual data, interview your people and return a costed plan, and that depth cannot be delivered through a questionnaire.
Independent boutique assessments typically run from around $15,000 to $75,000 and take two to eight weeks, according to Elevated Signal’s price analysis. CabinCo’s 2026 cost breakdown lands on the same range and the same underlying pattern, price tracks the complexity of your data and systems rather than the brand on the proposal. Below that sit lighter fixed-fee diagnostics for smaller firms, commonly around $2,500 over one to two weeks. Published sterling figures are thin, but UK boutiques doing serious evidence-gathering tend to quote in the low five figures.
Big-firm assessments from the likes of Deloitte, PwC and Accenture typically start above $100,000 and run for months, with hundreds of stakeholder hours attached. That depth makes sense for a regulated, multi-entity business that needs board consensus and deep compliance analysis. For a firm of fifteen people it buys thoroughness the business cannot absorb, and the stakeholder time alone can cost more than a whole boutique engagement.
When is a free or vendor-tied assessment enough?
A free self-assessment is enough when you have no defined use case yet and the stakes are low. Tools from Microsoft, AWS and others score you across their pillars in under an hour, surface obvious gaps in data or skills, and give your team a shared language for the conversation. Treat the result as a starting point rather than a plan.
Plenty of owner-managed firms are still at that orientation stage. YouGov polling of smaller UK firms found 31 per cent already using AI tools and a further 15 per cent planning to, with the rest watching and waiting. OECD analysis of the same population finds smaller firms commonly start with narrow, low-risk applications and lack a systematic view of their data, infrastructure and governance. If that describes you, a free tool is a cheap way to build the map before paying anyone.
The catch is in the incentives. Microsoft’s self-assessment scores you across its seven pillars, AWS partners run readiness workshops through its marketplace, and both are designed to surface opportunities to adopt that vendor’s platform. The inputs are also self-reported, so the output reflects how you think the business runs rather than how it does. Used knowingly, that is fine. The moment you are choosing between spending real money and not spending it, you need evidence someone gathered rather than answers you supplied.
What does it cost to choose the wrong provider?
The fee is the smaller loss. A weak assessment costs you the months you spend acting on it, the staff hours it consumed, and the confidence of a team that watched a consultant arrive, score the business and leave nothing behind that anyone could build from. Poor advice on governance can also leave you exposed under UK data protection law.
The pattern to watch for is the assessment as a loss-leader. Some providers structure cheap diagnostics as extended paid discovery, eight to twelve weeks with no named deliverable, or offer credit-back arrangements where the audit fee is rebated against future build work. There is nothing wrong with a provider hoping to win the implementation. A rebate, though, ties their findings to their pipeline, and the recommendations tend to drift towards whatever they happen to sell. Sub-£2,000 assessments marketed at smaller firms are frequently questionnaires with templated output and no inspection of anything.
The subtler failure is the vanity maturity score, a rating of “emerging” or “developing” against a generic framework, with recommendations attached to no process, owner or budget. Elevated Signal calls this the single most expensive mistake in the market, and Accenture’s own cross-industry research finds only a minority of organisations ever move beyond pilots into embedded, governed use of AI. A score does nothing to move you across that line. Neither does a provider who waves away data protection. If a proposed use case touches decisions about individuals, the UK GDPR’s automated decision-making rules apply, and an assessor who cannot discuss Article 22 in plain terms is a liability rather than a shortcut.
What should you ask before you decide?
Four questions in a single scoping call will separate a provider who gathers evidence from one who sells scores. Ask what evidence they collect, who they will interview, what the output commits you to, and what happens in the 90 days after the report lands. The answers reveal the method, and the method predicts the deliverable.
In the call, put them plainly and listen for specifics:
- What evidence do you gather beyond a questionnaire? You want to hear data samples, system walkthroughs and process mapping, with named steps rather than a proprietary scoring tool.
- Who will you interview, and how much of their time do you need? An assessment that only hears from the owner misses the manual work the owner cannot see.
- What does the output commit us to? A prioritised, costed roadmap with named owners is the right answer. A score with generic recommendations attached is the warning sign.
- What happens in the 90 days after the report? Some providers offer follow-up reviews or help with pilot design, others treat the report as the end. Either is workable, provided you know which you are buying.
Two more are worth the extra minute. Ask how the assessment handles data protection and AI governance, and ask whether you are free to take the report and implement it with someone else. A provider who is comfortable being hired for the diagnosis alone is the one whose diagnosis you can trust. The same discipline applies here as with any other supplier, and the due diligence checks for an AI supplier or startup transfer directly.
Whoever you hire, the assessment earns its fee in the 90 days after it lands. The right one tells you, with evidence, which two or three things to do next, what to leave alone for now, and how you will know whether the choice paid off. Buy that, from whoever genuinely offers it.
