How to Assess AI Readiness: A Six-Dimension Scorecard for UK SMEs

Why readiness matters more than enthusiasm

The fastest way to waste money on AI in 2026 is to start with the technology. Almost every failed UK AI programme we're called in to recover has the same root cause: enthusiasm outran readiness. Models were chosen before use cases were defined, use cases were defined before data quality was understood, and data was used before governance was in place.

A good readiness assessment forces the conversation in the right order: business outcome, then use case, then data, then governance, then platform.

The six dimensions

We score every readiness assessment across six dimensions. Each is rated 1 (none) to 5 (mature).

1. Strategic intent

Is there a named executive sponsor? Is AI tied to specific business outcomes – revenue, cost, risk, experience – with quantified targets? Is there a 12–24 month roadmap that the board has signed off?

2. Data foundations

Do you know where your critical data lives? Is it accessible, reasonably clean, and governed? Are you confident enough in lineage and quality to expose it to a model? Most UK organisations score 2 or 3 here – the data work is usually the long pole.

3. Governance and risk

Do you have an AI policy? An approved use list and prohibited use list? Model risk management, bias testing, human-in-the-loop controls? Alignment with the ICO's guidance and (where relevant) the EU AI Act? A plan for the upcoming UK AI legislation?

4. Skills and operating model

Do you have data engineers, ML engineers, prompt engineers and product owners – or partners who do? Is there a clear operating model for who proposes, approves, builds, deploys and monitors AI?

5. Infrastructure and tooling

Cloud platform decisions made? Identity, networking and secrets management ready for AI workloads? A position on commercial models (OpenAI, Anthropic, Google, Mistral) versus open-weight models? A view on cost controls?

6. Culture and change

Are staff being trained on AI use? Is there an internal community of practice? Is there honest leadership communication about job impact? A way to surface and resolve fears as much as opportunities?

Reading the scorecard

Add the six scores together. The total tells you what to do next:

• 6–12: Foundational. Don't deploy AI yet. Spend 6–12 months on data, governance and a single contained pilot.
• 13–18: Emerging. Deploy two or three controlled pilots in well-bounded use cases. Use them to build the operating model.
• 19–24: Operating. Scale pilots into production, formalise governance, start building or buying ML capability in-house.
• 25–30: Mature. Industrialise. Multiple production AI products, full lifecycle management, demonstrable ROI.

The five most common gaps in UK organisations

1. No executive sponsor. AI is being driven from IT or innovation, not the exec.
2. Data quality is unmeasured. No baseline, no monitoring, no SLAs on the data feeding the model.
3. No AI policy. Staff using ChatGPT, Copilot, Claude with no agreed rules.
4. No use case prioritisation. 30 ideas, no scoring, no chosen first three.
5. No measurement plan. Pilots launched without success criteria, so they can never be declared a success.

How we run the assessment

Our four-week AI readiness assessment combines stakeholder interviews, document review, technical environment review and a structured workshop. You leave with a scored scorecard, prioritised gap list, candidate use case shortlist, and a costed 12-month roadmap. Most clients use it as the foundation paper for board approval of their AI investment.

It's typically the first piece of work we do alongside our AI Strategy and Data Strategy engagements.