Score 15 statements from 0 to 2 across five areas: a clear job for AI, your information and accounts, people and time, safeguards, and money. It takes about 20 minutes. A total of 24 to 30 means pilot now, 16 to 23 means fix your zeros first, and 15 or less means groundwork before spending anything.
Score honestly and ask for evidence: a 2 means you could show someone proof in under a minute, not that you're fairly sure. Three answers override the total: no named job, nobody checking what customers see, and health or children's data you haven't mapped. Fix those first, whatever you scored overall.
How to score it in 20 minutes
Speed comes from having the evidence open before you start, so you're checking rather than remembering. Here's the timing:
- Two minutes to set up. Open your email, the admin page of your main software, your diary and last month's bank statement.
- Three minutes per area, five areas. Read each statement, look for the evidence, write 0, 1 or 2.
- Three minutes to total it and check the three override answers.
The scoring rule is the same for every statement: 0 means no; 1 means partly, or "I think so" without proof; 2 means yes, and you can show it. If you can, ask the person who does the day-to-day work to score it separately, then compare. Where you disagree by two points, the disagreement is itself a finding, usually that the owner thinks something is written down or set up that isn't.
Area 1: a job worth giving to AI
- You can name one task you want AI to help with, and roughly how many hours a week it takes. Vague goals ("we should use AI more") produce vague results, and the hours set the ceiling on what you can save. Score 2 if you can write a line like "drafting replies to booking enquiries, about 4 hours a week, done by the front desk".
- The task happens at least weekly, and the inputs look broadly alike each time. AI earns its keep on repeated, similar work; a monthly job where every case is different rarely repays the set-up. Score 2 if you can open five recent examples that follow the same pattern.
- You know how the task performs today, or can find out within a week. Time per item, volume, turnaround or how often it's redone. Without a starting number you'll never know whether AI helped. Score 2 if a figure is written down with a date; setting a baseline before you introduce AI shows how to get one from records you already keep.
Area 2: your information and accounts
- The information the task needs lives in one or two systems you control and can export. An assistant can only work from what it can see. Score 2 if you can export last month's bookings, orders or enquiries to a spreadsheet in five minutes, now.
- Your core accounts belong to the business, and the owner holds admin access. Email, files and your main business software. If the admin login sits with one employee or an outside helper, every AI set-up depends on them. Score 2 only if you can open the admin page yourself right now.
- You know which of your data is sensitive, and where it's kept. Customer contact details, payment details, health information, children's details. Score 2 if you can list each type and the systems that hold it.
Statement 5 is the one most often over-scored. A common version: the business email is on the company's own domain, so the owner scores it 2, but the domain and the admin login were set up by a web designer years ago and are still in their name. Try to log in to the admin page before you score, not after.
Area 3: people and their time
- A named person can give about two hours a week for six weeks. Every pilot needs someone to set it up, check the output and adjust it. Score 2 if the time is already blocked in their diary, not just agreed in principle.
- The people who do the task have been asked, not told, and are willing to try. A tool imposed on a sceptical team gets quietly ignored. Score 2 if you've had the conversation and they've said yes.
- Someone on the team is comfortable learning new software from help pages. Not technical, just willing to read, try and retry. Score 2 if they've set up a new app recently without help.
Area 4: safeguards
- You have a short written rule on what staff may and may not paste into AI tools. One page is enough. Score 2 if it exists and staff have read it; writing an AI usage policy gives you wording to start from.
- A person will check anything customer-facing before it goes out. Replies, quotes, posts, reminders. Score 2 if a named person does the check and it's part of the written process.
- The AI you'd use won't train on your business data, and nobody uses personal accounts for work. Business plans such as ChatGPT Business, Claude Team, Microsoft 365 Copilot and Gemini in Workspace don't train on business content by default; consumer plans rely on each person switching off the model-training setting in their privacy settings. Score 2 if you've checked the plan and work happens in one company account.
Area 5: money and timing
- You can fund the tools for three months without strain. For reference, ChatGPT Business Standard and Claude Team Standard both list at $25 a seat a month on monthly billing, with a two-seat minimum, and Microsoft 365 Copilot Business is $21 a user a month on an annual commitment. Score 2 if a monthly figure is in the budget.
- You'd be willing to stop after six weeks if it isn't working. Owners who can't stop keep paying for tools nobody uses. Score 2 if you'd say that to the team out loud.
- No other big change is under way in the next two months. A new till or booking system, a move, a peak season or a key person's leave all compete for the same attention. Score 2 if the next eight weeks look normal.
Reading the total, and the answers that override it
| Total | What it means | Do next |
|---|---|---|
| 24 to 30 | Ready to pilot one job | Pick the job from statement 1 and run a four to six-week pilot |
| 16 to 23 | Ready once the zeros are fixed | Fix every 0, most of which take days, then re-score |
| 0 to 15 | Groundwork first | Work through the fixes below before paying for any AI tool |
Three answers override the total, however high it is:
- Statement 1 scores 0. No job, no pilot. Everything else is preparation for something undefined.
- Statement 11 scores 0 and the job is customer-facing. Keep the pilot internal until someone is named to check.
- Statement 6 scores 0 and the job involves health or children's information. Map the data first, and take the question of what's allowed to your data-protection adviser.
Here's the whole thing as a scoresheet to print or paste into a document:
AI READINESS SCORESHEET Date: ________
0 = no 1 = partly / not proven 2 = yes, and I can show it
THE JOB
1 Named task + hours per week ...................... [ ]
2 Weekly, similar inputs ........................... [ ]
3 Today's numbers known (or within a week) ......... [ ]
INFORMATION AND ACCOUNTS
4 Data in 1-2 systems we control, exportable ....... [ ]
5 Business accounts, owner holds admin ............. [ ]
6 Sensitive data listed, locations known ........... [ ]
PEOPLE AND TIME
7 Named person, 2 hrs/week for 6 weeks ............. [ ]
8 The people doing the task are willing ............ [ ]
9 Someone learns software from help pages .......... [ ]
SAFEGUARDS
10 Written rule on what goes into AI tools .......... [ ]
11 Named person checks customer-facing output ....... [ ]
12 Business plan, no training, no personal accounts . [ ]
MONEY AND TIMING
13 Three months of tool costs budgeted .............. [ ]
14 Willing to stop at six weeks ..................... [ ]
15 No other big change in the next 8 weeks .......... [ ]
TOTAL ____ / 30
Overrides: 1 = 0? 11 = 0 (customer-facing)? 6 = 0 (health/children)?An optician, a music teacher and a pet shop, scored honestly
These three illustrative businesses show how differently the same sheet reads. Each scored against the one job it wanted AI for.
| Statement | Optician (6 staff) | Music teacher (solo) | Pet shop (2 shops, 9 staff) |
|---|---|---|---|
| 1 Named task and hours | 2 | 2 | 0 |
| 2 Weekly, similar inputs | 2 | 2 | 1 |
| 3 Today's numbers | 1 | 1 | 0 |
| 4 Exportable data | 1 | 2 | 1 |
| 5 Owner holds admin | 1 | 2 | 0 |
| 6 Sensitive data mapped | 2 | 1 | 1 |
| 7 Named person's time | 1 | 1 | 0 |
| 8 Willing team | 2 | 2 | 1 |
| 9 Learns from help pages | 1 | 2 | 2 |
| 10 Written AI rule | 0 | 1 | 0 |
| 11 Named checker | 2 | 2 | 1 |
| 12 Business plan only | 0 | 1 | 1 |
| 13 Budget | 2 | 2 | 1 |
| 14 Willing to stop | 2 | 2 | 2 |
| 15 Quiet eight weeks | 2 | 2 | 0 |
| Total | 21 | 25 | 11 |
The optician: 21, and two zeros to clear
The practice wanted help drafting recall reminders and replies to routine messages about collections and appointment changes. Its job, team and budget were solid, but it had no written AI rule, and two staff were already pasting message drafts into personal chatbot accounts. Because the practice handles health information and had mapped where it lives (statement 6 scored 2), there was no override. It spent a week on the two zeros: a one-page rule and one business workspace for the front desk. Re-scored, it reached 25 and started a pilot on collection messages, which carry no clinical detail.
The music teacher: 25, ready with one caution
A solo teacher with about 45 pupils wanted help turning lesson notes into short practice summaries for parents. Everything sits in a few accounts they control, the work is weekly and alike, and a quiet term lay ahead. The one caution was statement 6, which scored 1: pupils' notes include children's names and progress. The fix was a working habit rather than a project: first names only in anything pasted into the assistant, and the notes kept in the business's own account.
The pet shop: 11, with an override
The owner's aim was "AI for social posts, stock and customer questions", which is three jobs and therefore none, so statement 1 scored 0 and overrode everything. The online shop's admin login still sat with the web designer, nobody had time before the busy season, and there were no current numbers for any of the three jobs. The right answer was groundwork: recover the admin access, pick one job after the peak, and start counting. Choosing a first AI project is the step that turns statement 1 from 0 to 2.
Turning low scores into quick fixes
Most zeros take days to fix, not months. Here's what each usually needs:
| Low statement | Fix | Typical time |
|---|---|---|
| 1 No named job | Log where the week goes for five days, then pick the biggest repeated writing or admin task | 10 minutes a day, then 1 hour |
| 3 No numbers | Count last month's volume from email or your booking system, and time ten cases | 1 to 2 hours |
| 4 Data won't export | Ask the software vendor how to export, or choose a job whose data you can reach | 30 minutes |
| 5 Admin held elsewhere | Recover admin access or have the owner added as an administrator | 1 to 3 hours |
| 6 Sensitive data unmapped | List each type of personal data and the system that holds it | 45 minutes |
| 7 No one's time | Block two hours a week in one person's diary, and take something else off them | 15 minutes |
| 10 No AI rule | Write one page: what's allowed, what isn't, who to ask | 1 hour |
| 12 Personal accounts in use | Move work into one business account; until then, have staff switch off model training in privacy settings | Half a day |
| 15 Busy period ahead | Schedule the pilot for the quieter weeks after it | 5 minutes |
If several fixes apply, do them in this order: accounts and admin access, then the written rule, then the named person's time, then the job and its numbers. Access problems block everything else, and they're the cheapest to fix while nobody is waiting on them. Signs your business isn't ready for AI yet goes through each gap in more depth, with what to fix first.
Using an AI assistant to plan your fixes
Once you have a scoresheet, a chat assistant can turn it into a plan in a couple of minutes. A prompt like this works:
Here is my AI readiness scoresheet (0 = no, 1 = partly, 2 = yes).
[paste the 15 lines with scores]
Our business: [one line]. The job we want AI for: [one line].
Write a 30-day plan to raise every 0 and 1 that blocks a first
pilot. For each fix give: the task, who should do it, the time it
takes, and how I'll know it's done. Put fixes in the order they
should happen. Leave out anything that doesn't block the pilot.An illustrative reply for the optician, abridged:
"Week 1: Draft a one-page AI use rule (practice manager, 1 hour; done when all staff have signed it). Set up a shared AI workspace for the front desk (owner, 2 hours; done when both staff log in with work accounts). Week 2: Ask your practice software supplier how to export appointment data (practice manager, 30 minutes). Week 3: Build a dashboard of recall response rates…"
What you'd fix: the first two items are right and in the right order. The export question is sensible. The dashboard in week 3 is invented scope, since nothing on the sheet called for it, and the plan never mentions switching off model training on the two personal accounts in the meantime. Cut the extras, add the missing interim step, and the plan is ready. Assistants tend to add work; your job is to delete it.
What a 20-minute score can't see
A quick self-score is a filter, not a verdict. It won't tell you:
- Whether AI can actually do your job well. Only trying it on ten real, anonymised cases will show that, which takes about an hour.
- How your systems connect. Two tools that both "integrate with everything" may not pass the one field you need between them.
- What your contracts and obligations allow. Client terms, sector rules and data-protection duties need reading, not scoring.
- How clean your data really is. Statement 4 checks that you can export it, not that it's accurate.
If your total lands in the middle band and several systems or sensitive data are involved, an outside view can be worth paying for; what an AI readiness assessment involves explains what you'd get beyond this sheet.
Re-scoring after 30 days: what should have moved
Put a second 20-minute session in the diary a month after the first, and score it before looking at your old sheet. Three patterns are worth recognising:
- The zeros became 2s and the total reached 24 or more. Start the pilot. The groundwork is done and won't need redoing.
- The easy zeros moved but statement 7 didn't. This is the most common result, and it's the real finding: nobody has time. Until something comes off someone's plate, a pilot will stall at its first obstacle.
- Nothing moved. Readiness isn't the obstacle; priority is. That's a legitimate answer. Keep using a business AI account for your own low-risk drafting, and re-score when the business has room.
Keep each dated scoresheet. Six months on, the difference between the first and the latest is the clearest record you'll have of whether the business actually got ready or just talked about it.
Questions about scoring your AI readiness
Can I score one team or site separately?
Yes, and it's often more useful. Readiness varies inside a business: the office may have clean records and willing staff while the shop floor has neither. Score each team against the job you'd give AI in that team. Start your first pilot where the score is highest, because an early success in one team does more to win round the others than any presentation.
Does a high score mean I should buy an AI tool?
No. It means you're ready to pilot one job, and the tool may already be in your software. Microsoft 365 business plans include Copilot Chat at no extra cost, and Google Workspace plans now include Gemini. Try the job with what you have for two or three weeks, then buy only if the included features clearly fall short.
What if we score low but competitors already use AI?
Do the groundwork anyway; most of it takes weeks, not months. Recovering admin logins, writing a one-page AI rule and measuring one task are quick. Meanwhile you can still use a business AI account yourself for low-risk drafting, such as internal notes and first drafts of adverts, as long as a person checks anything customers will see.
Is this the same as a paid AI readiness assessment?
No. This is a quick self-score from what you can see. A paid assessment usually adds interviews with staff, a look at samples of your real data, a review of your systems and contracts, and a written plan. It's worth paying for when your score is borderline, several systems are involved, or the job touches sensitive data.
Further reads
- AI Readiness Checklist for Accountants, Solicitors, Consultants — A stricter version for firms bound by client confidentiality.
- Is Your Business Data Ready for AI? A Clean-Up Checklist — Go deeper on area 2 with a 50-record data test.
- How to Tell If a Process Is Ready to Automate With AI — Score the single process you picked, step by step.
- Microsoft 365 Copilot Readiness Checklist for Small Businesses — Extra checks if Copilot is the tool you'd start with.
- How Much Does an AI Readiness Assessment Cost? — What an outside assessment costs if your score is borderline.
- Shadow AI: Is Your Team Using AI Without Telling You? — Find out what staff already use before you score statement 12.
- Do I Need AI in My Business? A Decision Guide for Owners — A decision guide for owners: the three situations where AI is worth it, when waiting is sensible, a ten-question score and a 30-day test.
- Course, 1:1 Call, Audit, or Project: Which AI Help Do You Need? — What a course, a 1:1 call, an audit and a project each leave you with, five questions to find your starting point, and one deli's route through three.
- AI Audit vs AI Readiness Assessment: What's the Difference? — How an AI audit differs from a readiness assessment, the four things sold as an 'AI audit', worked barber and dry cleaner examples, and a DIY version of each.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI, Anthropic, Microsoft and Google business plan pages (seat prices, seat minimums, training defaults); Microsoft 365 business plan pages (Copilot Chat included); Google Workspace pricing page (Gemini included).