An AI readiness assessment tells you whether your business can get value from AI now, and what to fix first if it can't. The checklist below has 25 questions across seven areas: strategy and ownership, processes, data, systems and integration, people and skills, governance and risk, and money. Score each question 0, 1 or 2. Out of 50, 35 or more means you are ready to scale, 20 to 34 means you are ready for one contained first build, and below 20 means you should fix ownership, one process and its data before spending on tools.
I have one strong opinion about readiness, and it shapes the whole list. Readiness is a property of a process, not of a company. Nobody is "ready for AI" in general. A business can be ready to automate supplier invoice coding next month and years away from using AI in pricing. So answer the process, data and systems questions for one specific process: the one you would most like to fix. Answer the rest for the business as a whole.
Readiness is a property of a process, not of a company.
Is my business ready for AI? What the UK numbers say
Most UK businesses are earlier than the headlines suggest. The ONS reported in July 2026 that around 35% of businesses with ten or more employees now use AI, up from around 12% in late 2023, but only 15% say more than half their staff use it in daily work, and only 11% say more than half their workforce has had AI-related training.
The government's own survey points the same way. DSIT's AI Adoption Research, published in January 2026 from 3,500 interviews with UK businesses of five or more employees, found 16% currently using at least one AI technology. The most common reasons for not adopting were not having identified a use for AI (71%) and limited AI skills (60%). Among businesses already using AI, 54% felt ready to scale it, and only 13% described themselves as completely ready.
Meanwhile staff are not waiting. Microsoft's UK research in October 2025, a Censuswide survey of 2,003 employees, found 71% had used unapproved consumer AI tools at work and 51% did so weekly; 28% said their employer offered no approved option. In most businesses I look at, AI use is ahead of the policy and the training, and behind on anything that changes a process end to end.
How to use this AI readiness checklist
Get two people in a room for an hour: whoever runs the business and whoever owns the process you most want to fix. Answer honestly. Score each question:
- 2: yes, and you could show evidence (a document, a number, a named person).
- 1: partly, or yes but only in someone's head.
- 0: no, or you don't know.
"Don't know" scores zero. That is deliberate. Not knowing where your data lives is a finding. A fuller version, with the evidence behind every score, a shadow AI staff survey and a 90-day plan for each band, is in our AI readiness scorecard.
Strategy and ownership
The most common failure I see in AI projects is not technical. It is a project nobody owns, aimed at a problem nobody measured. These four questions test for that, and the first two are gate questions: a zero on either should be fixed before anything else.
| # | Question | Ready looks like | Not ready looks like |
|---|---|---|---|
| 1 | Can you name the one or two business problems you want AI to help with, in numbers? | "Quotes take three days to turn round and we lose about one job in five to faster competitors." | "We want to be more efficient" or "we should be doing something with AI". |
| 2 | Is one named person accountable for AI outcomes, with time set aside? | A director with half a day a week in their diary and the authority to change a process. | "IT is looking into it", or a steering group with no single owner. |
| 3 | Does the leadership team agree what AI is for here: cost, capacity, quality or product? | One sentence the CEO, CFO and operations lead would all recognise. | The CFO wants headcount savings, sales wants a chatbot, and nobody has noticed they disagree. |
| 4 | Have you agreed how you will measure success before you start? | A baseline: hours, volumes, errors or turnaround time, taken now. | "We'll know it's working when we see it." |
Processes
AI pays back on work that is repetitive, frequent and done by several people. It struggles with work that is different every time, or that each person does their own way.
| # | Question | Ready looks like | Not ready looks like |
|---|---|---|---|
| 5 | Have you found repetitive tasks done at volume, weekly or more often, by several people? | "The accounts team codes about 2,000 supplier invoices a month." | A list of one-off projects, or tasks done once a quarter. |
| 6 | Is the priority process done the same way whoever does it? | A written procedure, or steps any two people would describe the same way. | Three people, three methods, and an exceptions list only one of them knows. |
| 7 | Do you know how long the process takes and how often it goes wrong? | Rough time per item and an error or rework rate, even if estimated from a week's sample. | No idea, and no way to find out without asking around. |
| 8 | Is someone who does the work willing to help redesign it? | A team member who wants the tedious part gone and will test early versions. | The team hears about the project when it goes live. |
Data readiness for AI
This is the section where good ideas go to die, and it is better that they die here than four weeks into a build. DSIT's research found 21% of businesses cited data complexity (both too little data and unstructured data) as a barrier to adoption, and one mid-sized construction firm put it plainly to the researchers: "We've got too many various systems that don't all link together."
Answer these for the one process you picked. Question 10 is a gate question.
| # | Question | Ready looks like | Not ready looks like |
|---|---|---|---|
| 9 | Do you know which systems hold the data the priority process needs? | "Enquiries arrive in a shared mailbox, job history is in the ERP, rates are in a spreadsheet on SharePoint." | "Somewhere in the CRM, probably." |
| 10 | Can that data be reached without someone running a manual export? | An API, a connector, or a database someone can query. | Every month someone exports a CSV, fixes it by hand and emails it on. |
| 11 | Is it good enough to trust: do two systems agree on the same figure? | Customer counts and revenue match between finance and the CRM, or the differences are known and explained. | Every meeting starts with an argument about whose number is right. |
| 12 | Does each important dataset have an owner responsible for its accuracy? | A named person fixes errors at source. | Errors are fixed downstream in spreadsheets, differently by each team. |
| 13 | Do you know which of it is personal, special category or contractually confidential? | Data is classified (a simple green, amber, red scheme is enough) and staff know the rules. | Nobody has checked whether the client contract restricts where its data can go. |
The data questions are the ones businesses most often overrate. If you scored 2 on question 11 without checking, check.
Systems and integration
You do not need modern systems to start, but you need systems someone can connect to, and someone with the access to do it.
| # | Question | Ready looks like | Not ready looks like |
|---|---|---|---|
| 14 | Are the core systems for the priority process current, supported and able to connect to other tools? | Cloud or current on-premise versions with documented APIs or standard connectors. | A Sage or Access database that nobody has upgraded in years, with no supported way in. |
| 15 | Does someone inside the business administer those systems and can grant access? | An internal or outsourced IT lead who can set up a service account in a day. | Admin access sits with a former employee, or a supplier who takes three weeks to respond. |
| 16 | Do you know what AI features you already pay for in existing software? | A list of AI features in your Microsoft 365, CRM, finance and HR systems, and whether they are switched on. | Paying for Copilot licences nobody uses while buying a separate tool to do the same job. |
Question 16 is often the quickest win on the list. If you suspect you are paying for licences nobody uses, the one-week licence audit is a good companion exercise.
People and skills
Limited skills was the second most common barrier in the DSIT research, and the most common one among businesses already using AI. Training is cheap compared with a failed build. Skipping it is the usual reason a working tool goes unused.
| # | Question | Ready looks like | Not ready looks like |
|---|---|---|---|
| 17 | Do you know which AI tools staff already use, including personal accounts? | A short anonymous survey has been run and the answers are written down. | Leadership assumes nobody uses AI because nothing has been approved. |
| 18 | Has more than half your workforce had practical AI training on their own work? | Hands-on sessions where people built or used something on a real task. | A webinar, a link to a vendor's videos, or nothing. |
| 19 | Do the managers of the affected teams have the time and willingness to change how their team works? | The team lead has agreed to protect time for testing and adoption. | The team lead is at capacity and sees the project as head office's idea. |
On question 18, be strict. A lecture on prompting scores 1 at most. I've set out what training that changes behaviour looks like in how to train staff on AI.
Governance and risk
Governance should be proportionate. A 40-person firm needs a two-page policy, an approved tool and a training record, not a governance board. Question 20 is the last gate question.
| # | Question | Ready looks like | Not ready looks like |
|---|---|---|---|
| 20 | Do you have a written AI policy that people have been trained on? | A short policy with data rules, signed off by the CEO, covered in training, with a review date. | No policy, or a long one nobody has read that bans what people do every day. |
| 21 | Is there an approved tool list, with business accounts rather than personal ones? | Named tools, business plans with suitable data protection terms, accounts issued by the company. | Staff use free personal accounts because there is no alternative. |
| 22 | Would you know what to do if confidential or personal data went into an AI tool by mistake? | A reporting route, a named data protection lead, and a decision on whether to notify the ICO. | It would be found out by accident, if at all. |
| 23 | Do you know whether any planned use involves decisions about people, regulated advice or EU customers? | Each use is checked; anything touching hiring, credit or individual pricing gets a data protection impact assessment. | An AI CV-screening tool was switched on by the recruitment team without anyone asking. |
If you have no policy, my free AI acceptable use policy template can be adapted in a week, and the annotated version has drafting notes on every clause.
There is also the government's own governance self-check. DSIT published a draft self-assessment called AI Management Essentials for consultation in November 2024. Its December 2025 consultation response said feedback showed one tool struggled to serve organisations of different sizes, and that the government would instead develop future guidance on foundational governance aimed at SMEs. So it is not a finished readiness tool. The draft's questions on internal processes, risk and communication are still a reasonable second pass on questions 20 to 23 if you build or sell AI yourself.
Money
The last two questions are where many readiness efforts stall quietly. DSIT found that among businesses planning to adopt AI in the next year, 47% had no specific budget set and a further 25% didn't know what was budgeted.
| # | Question | Ready looks like | Not ready looks like |
|---|---|---|---|
| 24 | Is there a budget, even a small one, set aside for a first build and its running costs? | An agreed amount for the build plus a monthly line for software, model usage and upkeep. | "We'll find the money if the business case is good enough", with no one deciding what good enough means. |
| 25 | Do you know what an hour of the affected team's time costs, so you can work out payback? | A fully loaded hourly cost for the team, agreed with finance. | Savings will be argued about after the fact. |
How to score your AI readiness assessment
Add up the 25 scores for a total out of 50. Then look at three things, in this order.
Gate questions first. Questions 1, 2, 10 and 20 (a named problem, a named owner, reachable data and a policy) are the ones I'd treat as blockers. A zero on any of them means fix that first, whatever the total says. A business scoring 38 with no owner will still produce a pilot that stalls.
Then the lowest section. Work out each section's score as a percentage of its maximum. The weakest section is where your first effort goes. Most businesses I see are weakest on data or people. Very few are weakest on strategy, because everyone can name a problem; they just haven't measured it.
Then the total.
| Score | What it means | What to do next |
|---|---|---|
| 0 to 19 | Not ready to build. Something basic is missing: an owner, a measured problem or reachable data. | Name an owner, pick one process, take a baseline, approve one business AI tool and write a short policy. Repeat the checklist in two to three months. |
| 20 to 34 | Ready for one contained first build on your best-prepared process. | Pick the process that scored highest on questions 5 to 16, fix its specific data gaps, and run a fixed-scope build with a measured before and after. Train the team that will use it. |
| 35 to 50 | Ready to scale beyond one process. | Build a ranked, costed list of opportunities across teams and run two or three builds in parallel, each with its own owner and baseline. |
A word on honesty. Self-assessments flatter. In my experience the optimism concentrates on data and training, the two areas hardest to check from a meeting room. If your total lands near a boundary, assume the lower band.
What to do with the score
If you scored under 20, you don't need a consultant yet. You need a decision about who owns this, an afternoon choosing one process and a week measuring it. That costs nothing but time.
If you scored 20 or more, the next step is to turn "ready" into a specific first build: which process, what it will cost to build and run, and when it pays back. That is the gap between a readiness checklist and an AI audit, which checks your answers against the systems and the people doing the work and ends with a ranked, costed plan. I've set out what a good one includes, what UK providers charge and the red flags in what an AI audit is and what you should get for your money.
Whatever the score, don't skip the process questions to get to the tools. The reasons AI pilots fail are almost all visible in this checklist before anything is bought.
If you would like an outside view that tests these answers and ends with a first build scoped and priced, that is what our AI audit does, over two to three weeks.
Questions
How long does an AI readiness assessment take?
Done as a self-assessment with this checklist, about two hours: one hour for the leadership team to answer the strategy, people, governance and money questions, and another hour with the person who owns the priority process to answer the process, data and systems questions. An externally run readiness assessment or AI audit usually takes one to three weeks, because it checks the answers against the systems and the people doing the work.
What is data readiness for AI?
Data readiness for AI means the information a specific AI use needs can be found, reached without manual exports, trusted, and used lawfully. In practice: you know which system holds it, there is an API or connector, two systems agree on the same figure, someone owns its accuracy, and you know whether it includes personal or confidential data. It is judged per use case. A business can be ready for one process and nowhere near ready for another.
Is there a free government AI readiness tool in the UK?
DSIT consulted on a free self-assessment called AI Management Essentials from November 2024. Its December 2025 consultation response said the government would refine the approach and develop future guidance focused on foundational governance for SMEs, so it is not a finished, general readiness tool. Its draft questions are still a useful check on governance. For commercial readiness, a structured checklist like this one, or Innovate UK Business Connect's adoption support, is more practical.
What is a good AI readiness score?
On this checklist, 35 or more out of 50 means you can run several AI projects in parallel. Twenty to 34 means you are ready for one contained first build on your best-prepared process. Below 20 means fix ownership, one process and its data before buying anything. The total matters less than the lowest section score and any zero on the four gate questions, which should be fixed first whatever the total says.
Should a small business do an AI readiness assessment?
Yes, but a short one. A business of ten to fifty people can answer most of these questions in an hour and will usually find that one process, one owner and one approved tool are the whole plan. The money and governance questions still matter at that size: a free consumer AI account and no data rules is the most common gap in small firms, and the cheapest one to fix.