An organisation is ready for AI when its data can be reached and trusted, someone owns the outcome, and staff know the rules. Tools come last. This guide gives you a 15-minute paper check of those foundations, explains what each result means, and shows where to go next.
Why readiness comes before tools
Most failed AI projects in small and mid-size organisations are not failures of the AI. They fail because the data the tool needed was spread across spreadsheets, inboxes and WhatsApp threads, because nobody owned the result, or because staff used the tool in ways nobody had agreed. A readiness check finds those gaps while they are still cheap to fix.
It also tells you what not to buy. A business with weak data foundations gets more from organising its records than from any AI subscription.
The 15-minute paper check
Score each statement 0 (not true), 1 (partly true) or 2 (fully true). Be honest: the score is for you.
| Area | Statement |
|---|---|
| Data | The records we would use (sales, customers, stock, finance) are in one place, not scattered across personal files. |
| Data | We trust those records: they are complete and kept up to date. |
| Access | We know who can see which data, and access is removed when someone leaves. |
| Security | Accounts use strong passwords and, where possible, two-step login. |
| Ownership | One named person is responsible for each AI or automation project. |
| Process | We can describe, step by step, the task we want AI to help with, and how long it takes today. |
| People | Staff have written rules on what they may and may not paste into AI tools. |
| Oversight | A person checks AI output before it reaches a customer or a decision. |
| Vendors | We know where our AI tools store our data, and we have read their terms. |
| Compliance | We know our obligations under the Data Protection Act, including whether we must register with the ODPC. |
Add up your score out of 20.
What your score means
- 15 to 20: ready to pilot. Your foundations hold. The risk now is spreading effort too thin: pick one task with a clear owner and a measured baseline, and run it for 30 days.
- 8 to 14: building foundations. You can start, but fix your two lowest areas first. A pilot built on missing data or unclear ownership tends to stall.
- 0 to 7: getting started. A normal place to begin. Put the basics in place, reachable data, access rules and a named owner, before you pay for AI tools.
These bands mirror the ones our full assessment uses: 70 and above out of 100 is "ready to scale AI", 40 to 69 is "building AI foundations", and under 40 is "getting started with AI".
Responsible use is scored on its own
A strong headline score can hide a weak spot in how AI is used. That is why the full assessment scores responsible use separately: privacy notices, consent, human review, where your data is stored, staff AI rules and ODPC registration. An organisation with excellent data but no rules on what staff paste into chatbots is not ready, whatever its headline score.
Who should take the check
Readiness looks different from different desks, so do not fill it in alone. Ask three or four people to score it separately: the owner or manager, whoever handles finance, whoever runs day-to-day operations, and whoever looks after your computers and accounts, even if that is an outside provider.
Then compare. Where everyone gives the same score, you have a shared picture. Where scores differ by two points, you have found the conversation worth having: the manager may believe the customer records are complete while the person who uses them every day knows they are not. Those gaps in understanding are often more useful than the total.
Keep the sheet. Repeating the check in six months, with the same people, shows whether the foundations are improving.
Turning the result into a plan
- List five tasks your team repeats every week: answering the same customer questions, chasing invoices, writing reports, checking stock.
- For each, note who owns it, how many hours it takes, and where its data lives.
- Cross out any task whose data you cannot reach today. AI cannot use data it cannot get at.
- Of what is left, circle the one with the most hours and the least risk if a draft is wrong. That is your first pilot.
- Agree how you will measure it: hours saved, errors avoided, or replies sent faster. Measure for 30 days before you expand.
Common mistakes
- Buying a tool before choosing the task. The tool should fit the task, not the other way round.
- Skipping the baseline. Without knowing how long a task takes today, you cannot show the pilot worked.
- No human check. AI drafts can be confidently wrong. A person should approve anything a customer sees.
- Treating readiness as a one-off. Re-check every year, or when you add a new AI tool.
Take the full assessment, free
The free AI Readiness Score asks 21 questions: 15 across nine areas, from data quality and access controls to vendor governance, incident response and skills, and six on responsible use, scored on their own. You get a score out of 100, a breakdown by area and a branded PDF report. A strong result (80 or more overall, with 70 or more for responsible use) earns a self-assessed badge you can display for 12 months.
This information is given as a guide. Confirm it before you use it, and adapt it to your own situation.
Common questions
How long does an AI readiness assessment take?
The paper check above takes about 15 minutes. Our full online assessment has 21 questions and gives you a score, a breakdown by area and a PDF report.
Do small businesses need an AI readiness assessment?
Yes, and they gain the most from one. A small business has less room for a failed project, and the check often shows that organising data or writing staff rules is the best first step.
What is a good AI readiness score?
On our 100-point scale, 70 or more means your foundations are ready to scale AI. What matters more is fixing your lowest areas, whatever the total.
Is the badge a certification?
No. It is self-assessed and says so, and it lapses after 12 months. It shows you have checked your foundations, not that an auditor has.
