AI Readiness Checklist: A 15-Minute Self-Assessment for UK SMEs

AI readiness is how well your business could put AI to work on one real task. It comes down to six things: a problem worth solving, data you can reach, willing users, a clear process, rules for safe use, and an owner. The checklist below scores each of them in about 15 minutes, for a total out of 36.

You do not need a digital strategy or technical staff to use it. We know of no official UK standard for AI readiness, so the framework is our own, written for UK SMEs. This article covers what readiness means, the 18 questions, how to read your score, a worked example with a fictional business, and what a paid assessment adds over doing this yourself.

Key takeaways

• AI readiness applies to a specific task. Score the process you have in mind, and score others separately.
• The checklist has 18 questions in six areas. Each scores 0, 1 or 2, giving a total out of 36 and one of four result bands.
• Two areas act as gates. Without a named problem, or without a rule for personal data, a high total does not mean go.
• A low score is a useful result. It shows which gap to close first, and sometimes that AI is not the next move.
• The checklist is free. An outside view from us is a one-week, £3,000 assessment, and the checklist is a sensible first step before it.

What does AI readiness mean for a small business?

AI readiness means your business has what it needs to get a working result from AI on a specific task, rather than a demo that never reaches daily use. You will also see it called an AI maturity assessment or an AI readiness framework. The questions overlap, though large-company versions weigh infrastructure and governance programmes that most SMEs do not have.

It is a different tool from the five-stage model in our AI transformation roadmap. The stages describe how far a process has come. This checklist tells you whether one process is ready to pilot, which is the move into stage 3.

The government’s research shows where businesses get stuck. DSIT’s AI adoption research asked all businesses what had stopped them adopting AI. The top barriers were “a lack of identified need for AI (71%) and limited AI skills, expertise and knowledge (60%)” (DSIT, AI adoption research, fieldwork February to May 2025). Those two barriers are why our checklist opens with a named problem and has a full section on people.

Use of AI is already common. In the ONS Business Insights and Conditions Survey, 30% of businesses reported using at least one type of AI technology in September 2026. The figure was 29% for businesses with fewer than 10 employees and 38% for those with 10 or more (ONS, 8 October 2026). Using a chat tool is a different thing from being ready to build AI into a process, and the second is what this checklist measures.

If you want a second opinion on method, Microsoft publishes an AI Readiness Wizard that asks 10 questions and then suggests the next area to focus on. Ours is longer and scores one specific process. If you would rather have an outside view, our AI readiness assessment is a one-week programme at £3,000.

How does this AI readiness checklist work?

Pick one process and score the checklist for it. Good candidates are first replies to enquiries, chasing late invoices, or sorting incoming documents. If you have several candidates, score all six areas again for each one. Some answers carry over, such as your written rule on AI tools, but whether the process touches personal data, who checks its output and who would own it do not. Area 5 is a gate, so never copy it across.

For each question, score 0 if the left-hand description fits, 2 if the right-hand description fits, and 1 if you are somewhere between. If you are unsure, score the lower number. Answer for the business as it is today, and ignore planned hires.

Do it with the person who does the work as well as the person paying for it. Their answers will differ, and the gap between them is information. Treat the framework as a thinking tool. It is not a standard, and a score is not a prediction.

The AI readiness checklist: 18 questions in six areas

Score each AI readiness question below, then add up each area. The maximum for an area is 6.

1. A problem worth solving

This area comes first because “no identified need” was the most common barrier in the DSIT research above. Without a named task, nothing else on the list matters.

Question	Score 0	Score 2
Can you name one task AI could help with?	Only “we should use AI”	One sentence, written down
Do you know what it costs today?	No idea	Hours a week or £ a month
Could a number show it got better?	No measure	A number we track now

2. Data and systems

AI works from the information you give it. When we run our own assessment, a data check is part of the week for this reason.

Question	Score 0	Score 2
Where does the information live?	Paper, inboxes, memory	A system someone owns
Can it be exported or connected?	Locked in, or on paper	Export or integration exists
Is it mostly accurate and current?	Often wrong or stale	Mostly right, and checked

3. People and skills

A system nobody uses saves nobody anything. These questions check who would use it and who would look after it.

Question	Score 0	Score 2
Do the people doing the work want help?	Not asked, or resistant	Asked, and keen
Has anyone used AI tools for work?	No one	Several, with some training
Would one person own it day to day?	No one has time	A named person with time

4. Process clarity

You can only hand over a task you can describe. This is the area where a spreadsheet or a person’s memory most often stands in for a process.

Question	Score 0	Score 2
Can you explain it in five steps?	Lives in one head	Written down, or easy to explain
Does it run the same way most times?	Different every time	Mostly the same
Do you know when a person must decide?	No	Exceptions are known

5. Risk and data protection

This area is a gate. A good total means little if you cannot say what personal data the process touches.

Question	Score 0	Score 2
Do you know if it touches personal data?	Not sure	Yes, and mapped
Is there a written rule for staff on AI tools?	No rule	Written, and staff know it
Does a person check output before customers see it?	Nothing is checked	A person approves it

Two public sources sit behind these questions. The ICO’s guidance on AI and data protection covers accountability and governance, transparency, lawfulness, accuracy, fairness, security and individual rights. The ICO says on that page that the guidance is under review because of the Data (Use and Access) Act, so check it for the current position.

A data protection impact assessment (DPIA) is required where processing is likely to result in a high risk to people’s rights and freedoms. The ICO’s list of processing that needs one includes innovative technologies, including AI, in combination with another criterion (ICO, when do we need a DPIA).

For a plain-English view of the security side, the NCSC’s AI and cyber security: what you need to know is written for managers, board members and senior executives. It covers risks such as hallucinations, prompt injection and data poisoning. This section summarises public guidance and is not legal advice.

6. Budget and ownership

AI projects stall when nobody can say yes, or nobody can say stop. These questions test both.

Question	Score 0	Score 2
Is there a sponsor who can say yes?	No	Named, and keen
Is money or time set aside?	None	Agreed for a first project
Would you know when to stop?	No review point	A date and a go or no-go number

How do I read my AI readiness score?

Add the six area totals to get your AI readiness score out of 36. Before you look at the band, check the two gates.

• Gate 1: if area 1 totals 2 or less, you do not yet have a problem worth solving, whatever your total. Name and measure one task first.
• Gate 2: if area 5 totals 2 or less, settle the rules for personal data and staff use of AI tools before anything touches customer data.

Score	Band	Next step
0 to 11	Not yet	Name and time one task
12 to 23	Groundwork first	Raise your lowest area
24 to 31	Ready to pilot	Scope one pilot
32 to 36	Ready to extend	Pick the best-payback process

0 to 11: Not yet

AI would be a tool looking for a job. What it can do now is modest: individual staff using a general assistant to draft and summarise, inside a written rule on what they may paste into it. Your one next step is to pick a repetitive task and time it for two weeks, so area 1 has numbers. Our guide to where automation pays off shows how to compare candidates.

12 to 23: Groundwork first

Some foundations exist and some are missing, which is a normal place to start. AI can already help in a contained way, such as drafting first replies, summarising documents or sorting inbound email, with a person approving every output. Your one next step is to raise your lowest-scoring area by two points before you buy anything. If that area is people, team training starts from £1,500 for up to 15 people.

24 to 31: Ready to pilot

You can run a scoped pilot on one process. AI can read incoming requests, pull the facts from your system, draft the response and pass exceptions to a person. That pattern is an AI agent. Your one next step is to write the pilot on a page: the task, the measure, the owner, the review date and the person who approves outputs. A first agent for one scoped job starts from £5,000 through our AI agents and automation service.

32 to 36: Ready to extend

Your risk shifts to overbuilding. AI can run several processes, with confidence thresholds deciding what goes ahead automatically and what waits for a person. SBU, a UK energy broker, had us build an automation layer across Salesforce, an LLM (Anthropic Claude) and its email. Replies are classified and responses drafted for approval, and an 85% confidence threshold decides what runs automatically. Within three months the team was generating five times the qualified leads, without adding headcount.

Your one next step is to choose the process with the best payback and plan the first 90 days using our AI transformation roadmap. Self-scores can run generous, so ask the person who does the work to challenge every 2.

A worked example: scoring a fictional business

This example is illustrative. It is a fictional 14-person letting agency, scoring the process of writing first replies to rental enquiries, which takes the team about 8 hours a week.

Area	Score	What drove it
1. Problem	5 of 6	Named task, rough hours only, reply time tracked
2. Data	3 of 6	Enquiries in a shared inbox, details in a CRM and spreadsheets
3. People	4 of 6	Team keen, one user of AI tools, owner has little time
4. Process	3 of 6	Mostly the same, exceptions not written down
5. Risk	2 of 6	Personal data involved, no written rule
6. Budget	3 of 6	Keen sponsor, some staff time, no stopping rule

The total is 20 of 36, which is the groundwork first band. Gate 2 also applies, because area 5 is 2. The agency should not start with a tool. It should write the staff rule, run a data protection check on the enquiry data, and document the exceptions, which lifts areas 4 and 5.

With those done, it could move to a pilot in which AI drafts first replies from the CRM record and a person approves each one. The measure is the one it already tracks: reply time. If the pilot misses it by the review date, it stops.

What does a paid readiness assessment add?

The checklist tells you where you stand. It cannot rank your processes by payback, put £ figures on them, or give you an outside view of your own data and workflows. That is what the paid version is for.

Our AI readiness assessment is a one-week programme at a fixed £3,000. It includes three calls with our most experienced leaders, your data, workflows and team skills mapped, and a research and proposal package to keep, with estimated ROIs. Where AI is not worth it, the proposal says so. There is no obligation to build anything with us afterwards.

If you do ask us to build the top opportunity, follow-on builds start from £8,000, with scope and price fixed before work starts and payment by milestone as each stage is delivered. All code and IP are yours once the work is delivered and paid for.

You may not need it. If you score 32 or more and already know the process, scoping a build directly is the cheaper route. If you score under 12, a paid assessment is early, and the two-week timing exercise in the first band is a better use of the money.

Your next step

Run the checklist today, and write down two things: your total and your lowest-scoring area. That area is your first piece of work.

If the result leaves you unsure which process to start with, or you want an outside check on your data, book our AI readiness assessment. When you are ready to plan the build, our AI transformation roadmap sets out a 90-day plan that starts with one process.