You're ready for AI when you can name one job you want it to do, point to where the right answers live, and say who picks up when it gets stuck. If you can't do all three yet, don't buy a tool; sort the process first, and I promise the AI part gets much easier.
In over five years of business analysis, I've found most "AI isn't working for us" problems don't come down to the model. They come down to the job, the sources or the handoff being fuzzy. So treat this page as a quick AI readiness assessment for small businesses: be honest, and fix what's missing before you pick anything.
What this test is really asking
Three things, really. What should the assistant answer? Which sources can it trust? Who takes over when it doesn't know? Mark it against how you run today, not how you'd like to.
If you can't name the job, or point to the material the assistant should use, you're not choosing a tool yet. You've got a process problem. Fix that first, then pick the assistant that fits.
This is the bit the hype skips. Demo chatbots look brilliant because someone gave them tidy answers. Does your business have those? Or does pricing live in one person's head? Put AI on top of that and it repeats the mess, faster.
You're in good company if this feels hard. When the ONS asked UK firms what was holding them back from AI in 2023, the biggest blocker wasn't cost. It was working out what to actually use it for (39% said so, against 21% for cost). That matches what I hear on calls almost every week.
Say you run a six-person lettings agency and want a chatbot on your website to answer tenants overnight. Then you go looking for the answers: the pet policy changes from landlord to landlord and lives in old email threads, the deposit questions sit in one negotiator's head, and the repairs process is "ring the property manager". A bot would fill those gaps with confident guesses, so the first job is a short written answer sheet per landlord and a named person for urgent repairs, not a chatbot.
Test yourself
Give yourself one point for each line that's already true. If it only lives in a slide or a Slack thread, it doesn't count.
- We can name one repeated question or task an assistant should handle every week
- The answers live in documents, tickets, or systems we own, not only in people's heads
- Someone can update those sources without waiting on a developer for every change
- We know who reviews odd or high-risk answers before they reach a customer
- We have a clear handoff when the assistant should stop and a person continues
- We can measure whether the assistant reduced wait time or repeat questions
- 0-2: Start with documentation and ownership.
- 3-4: You are close. Tighten sources and handoffs.
- 5-6: You are ready to scope a first assistant with a narrow job.
Be strict. "Sort of" is a no. A low score just tells you what to fix before you spend money.
What good readiness looks like
- One job, in plain language (for example: answer product questions from the handbook)
- A short list of sources the assistant is allowed to use
- A written rule for when it must say it does not know
- A named owner for keeping the content current, and a named owner for escalation
That's it. A small, clear job, material you own, and a person who takes over when the job ends.
One more thing before you switch anything on. If the assistant will see customer details, work out where that data goes. The ICO's guidance on AI and data protection is the place to start, though it's being reviewed after the Data (Use and Access) Act, so check it's current. I've also written about the decisions I'd never hand over to AI.
Picture a small accountancy practice in January, with clients ringing to ask what's still missing from their self assessment. A sensible first job is: answer "what do you still need from me?" using only the practice's records checklist, the engagement letter and each client's status in the practice management software. Anything that sounds like tax advice, a penalty or a deadline extension goes straight to a named accountant.
Where to start with Oduro
If you scored 5 or 6, start with an assistant grounded in your own material. Here's how we approach AI knowledge assistants. Keep the first version narrow enough that you can tell within a few weeks whether it's helping.
Scored lower? You're not behind. Government research by DSIT found that only about a third of businesses planning to use AI feel ready to do it. Often the smartest next step isn't AI at all. It's a simple automation or a tidier system, so have a look at what to automate first.
Either way, test any AI agent before customers see it. For more on where AI fits, browse our guides.
How the full assessment scores you
You rate about 20 statements from 1 (not at all) to 5 (fully in place). Each section (Processes, Data, Routine, People, Guardrails) gets a level from 1 to 4 from your average there: under 2 is level 1, under 3 level 2, under 4 level 3, and 4 or more level 4. Your overall average puts you in one of four stages:
- Firefighting: work still depends on who's in the room. Write down the jobs that stall before adding any tool.
- Foundations: some groundwork is there. Start with the statements you marked low.
- Building: the work is mostly visible. Join it up, name an owner and measure the time you get back.
- Ready to scale: your work is written, connected and measured enough to hand routine pieces to a system, with a person still deciding anything sensitive.
You see your overall score straight away. Add your name, email and company for the full breakdown, which we'll email you too.
Questions people ask
How do I know if my business is ready for AI?
When you can name one job for it, point to where the answers live, and name who takes over when it's stuck. The test above is a good first check.
What is an AI readiness assessment for small businesses?
It's a check on the groundwork AI depends on: your processes, data, routine work, people and guardrails. Ours takes about ten minutes.
What should I fix before using AI?
Write down how your key jobs get done, get client data into one place, and decide who owns the content. That pays off even without AI.
Is it safe to put customer data into an AI tool?
It depends on the tool and the data. Check your UK GDPR position first, starting with the ICO's guidance, and get proper advice if the data is sensitive.
Sources
- Management practices and the adoption of technology and artificial intelligence in UK firms: 2023 (Office for National Statistics)
- AI adoption research (Department for Science)
- Guidance on AI and data protection (Information Commissioner's Office)