Your business is ready for AI when you can point to a specific, repetitive piece of work that is written down, uses data you can actually reach, has a known cost when it goes wrong, and has one person who will own the result. Tools, budget and technical skills come after that. The ten questions below test for those conditions. Count your "yes" answers, and the score tells you what to do next.
This isn't an audit. It's a filter you can run yourself in about fifteen minutes before you speak to anyone, including us. If you want to see what a full audit involves, we've written that up separately in The AI-readiness audit, step by step.
The 10 questions
Answer each one with a plain yes or no. If you're not sure, count it as no.
| # | Question | Tests for |
|---|---|---|
| 1 | Can you name one workflow that happens at least weekly and takes real staff time? | A worthwhile target |
| 2 | Could someone write down its steps, from trigger to "done", in under a page? | Defined work |
| 3 | Is the information it uses already digital, in software you can export from or connect to? | Reachable data |
| 4 | Do the inputs follow a broadly consistent shape (same form, same fields, same kind of request)? | Automatable inputs |
| 5 | Do you know roughly what it costs today, in hours, delays or errors? | A baseline to measure against |
| 6 | Do you know what a mistake in this workflow would cost, and who would catch it? | Risk understanding |
| 7 | Is there a clear place where a person could approve the AI's work before it goes out? | A safe first design |
| 8 | Is there one named person who would own this working, not just launching? | Ownership |
| 9 | Do you have a written rule for what company or client data staff may put into AI tools? | Basic governance |
| 10 | Would the team using it be involved in designing it, not just told about it afterwards? | Adoption |
What does each question actually tell you?
1. A weekly, time-consuming workflow. AI pays off on repetition. A task done twice a year isn't worth automating however annoying it is. If you can't name a frequent one, start by keeping a list for a week of where time goes.
2. Written down in under a page. This is the question most businesses fail, and it's the most important one. If nobody can describe the steps, there's nothing concrete to build, and a project turns into discovery paid for at build rates. The fix is free: have the person who does the work write it down.
3. Digital and reachable data. An AI system can't read a filing cabinet or someone's memory. Information inside software with an export or an API can be worked with. Information in personal inboxes, paper or "ask Sarah" has to be moved first.
4. Consistent inputs. A standard intake form is easy. A mix of emails, phone notes and PDFs in ten layouts is still possible, but it costs more and needs more testing. That changes the price, not necessarily the answer.
5. A known baseline. If you don't know what the workflow costs today, you won't be able to tell whether the AI improved it. Even a rough figure ("about six hours a week across two people") is enough. We explain why this matters after launch in Is your AI actually working?
6. The cost of a mistake. A wrong draft that someone edits costs a minute. A wrong payment, a wrong medical instruction or a wrong legal filing costs far more. Knowing which kind of workflow you have decides how much the AI should do on its own.
7. A place for approval. The safest first project is one where the AI prepares and a person approves. If you can see where that checkpoint would go, you can start safely. We cover how to set that balance in How much should an AI decide on its own?
8. One owner. Projects without an owner stall after launch. Nobody notices when quality drops, and nobody updates the system when the business changes. The owner doesn't need to be technical. They need to care whether it works.
9. A data rule. Even a one-paragraph policy (which data may go into which tools, and what never leaves) prevents the most common early mistake. If you want a structured starting point, the U.S. National Institute of Standards and Technology publishes a voluntary AI Risk Management Framework, including a profile specific to generative AI. A small business doesn't need all of it, but it's a sound checklist to borrow from.
10. Involving the team. People who help design a tool tend to use it. People handed a tool tend to work around it. Asking the person who does the work what they'd change costs nothing.
How should you read your score?
| Your "yes" count | Where you are | What to do next |
|---|---|---|
| 8–10 | Ready to scope a first project | Pick the workflow from question 1 and get it scoped properly, with cost, effort and a definition of "done" |
| 5–7 | Close; there are specific gaps | Fix the "no" answers first. Most (2, 5, 8, 9, 10) cost nothing but a conversation and a document |
| 0–4 | Not yet, and that's useful to know | Don't buy anything yet. Start with questions 1 and 2: find the repetitive work and write it down |
A low score isn't a failure. It saves you from paying for a build before the groundwork exists. Most of the gaps this check finds are organisational, not technical, and you can close them without a vendor.
Which "no" answers matter most?
Not all ten are equal. Three "no" answers are worth stopping for:
- No to question 2 (not written down). Nothing else can happen until the work is defined.
- No to question 3 (data not reachable). The first project becomes a data project, which is fine, but it should be scoped and priced as one.
- No to question 8 (no owner). You can launch without one, but it won't last.
A "no" on question 4 or 6 doesn't rule you out. It points to a more careful design, usually with more human approval at the start.
What should you do with the result?
Whatever your score, the next step is the same: choose one workflow, not a strategy. A readiness score tells you whether you can start. Deciding which workflow to start with, and whether the right fix is software you buy, a simple automation or a custom build, is a separate question. We cover it in Buy, build, or automate.
If you scored 5 or more and want a second opinion on which workflow to start with, our audit is free. It's a 30-minute call plus a written roadmap ranked by return and effort, and the report is yours to keep whether or not you build with us. If a build follows, the costs are published on our pricing page.
Scored well and want to know where to start? Book a free audit.
Discuss a project with Abrams & Luchanski →