Most owners we talk to have already tried AI. They ran some prompts, saw something impressive, then couldn't turn it into anything the business actually uses. That's rarely a tool problem. It's usually that the business had nothing solid to hand the tool.
Readiness isn't about being technical. It's a short list of conditions you can check yourself: a repeatable process, findable data, enough volume, a named owner, room for error, and a real budget. What follows is a self-scoring check you can run at your desk in about 20 minutes. If you'd rather be asked the questions than score yourself, our free 12-question AI Readiness Score covers the same ground.
The Bottom Line
Why readiness matters more than picking the right tool
Small business AI use is no longer early. Adoption jumped 41% in a single year, from 39% in 2024 to 55% in 2025, according to a Thryv survey. The same survey put companies with 10 to 100 employees at 68%, up from 47%. A QuickBooks survey reported by the OECD lands on that same 68%, up from 48% in mid-2024.
Adoption and results are two different things. MIT's 2025 GenAI Divide report found that implementations done with an outside partner succeeded about 67% of the time, against roughly 33% for builds companies attempted on their own. We broke down what that gap means for whether you need outside help at all in a separate post.
The most common barrier is knowledge, not budget. 51% of business leaders identified insufficient knowledge about AI as their main barrier, per the Institute of Directors. Running a readiness check on your own business is the cheapest way to close part of that gap, because it tells you what to ask for.
This check needs no IT department and no framework. It asks whether the work you want to automate has the raw material an AI system needs. If you're fuzzy on what AI does day to day in a small business, start with the plain-English version.
How to run the check in 20 minutes
Pick one workflow before you start. Not "AI for the business." One thing: the quotes you write, the phone that rings unanswered, the invoices you chase, the report you rebuild every Monday morning.
Then work through the six dimensions below with these four rules:
This is a check, not a study. A longer version won't be a more honest one.
The six dimensions to score
1. Does the process already exist?
AI can copy a process. It can't invent one for you. If three people do this job three different ways, there's no standard for a system to hit and no way to tell whether the output is right.
What good looks like (4 to 5)
What not-ready looks like (1 to 2)
2. Can you find and export your data?
The test isn't whether your data is clean. It's whether you can name where it lives and get it out this week. AI can't price, decide, or answer from records you don't have.
What good looks like (4 to 5)
What not-ready looks like (1 to 2)
This one is the gate. Score it 1 or 2 and the total doesn't matter yet, as the score bands below explain.
3. Does it happen often enough for the math to work?
Volume decides payback, and it's the dimension owners skip most. Count how often it happens and how long it takes. Two hours twice a week is roughly 200 hours a year. Two hours twice a year is a rounding error.
What good looks like (4 to 5)
What not-ready looks like (1 to 2)
4. Is there one named owner?
Every stalled build we've seen was missing the same thing: a person whose job it was to answer questions during the build and accept the output afterward. A committee isn't an owner. Neither is "we'll figure that out later."
What good looks like (4 to 5)
What not-ready looks like (1 to 2)
5. What happens when the AI is wrong once in twenty?
Every system makes mistakes, including the manual one you run now. The real question is what a single wrong answer costs and who catches it before it leaves the building.
What good looks like (4 to 5)
What not-ready looks like (1 to 2)
6. Is the budget honest?
Market pricing for this work is knowable, so plan against it. A first fixed-scope build generally runs $2,500 to $10,000. Ongoing support runs $1,500 to $8,000 a month. Those are ranges, not a quote, and where you land depends on the workflow.
What good looks like (4 to 5)
What not-ready looks like (1 to 2)
Your scorecard
| Dimension | What it tests | Your score (1 to 5) |
|---|---|---|
| 1. Process | The task repeats the same way every time | |
| 2. Data | You can find it and export it | |
| 3. Volume | It happens often enough to pay back | |
| 4. Owner | One named person accepts the output | |
| 5. Error tolerance | A wrong answer gets caught and costs little | |
| 6. Budget | Build and support money is real | |
| Total | / 30 |
What your score actually means
One override before you read the bands. If dimension 2 scored 1 or 2, treat yourself as Foundation First regardless of the total. A big time drain with no usable records isn't a build, it's a data project wearing a build costume.
| Total | Band | What we'd tell you |
|---|---|---|
| 26 to 30 | Ready to build | Stop assessing and scope one workflow. You have the raw material. |
| 20 to 25 | Ready after one fix | Name your lowest dimension, fix it in the next 30 days, then start. |
| 13 to 19 | Foundation first | Usually process or data. Buying software now just automates the mess faster. |
| 6 to 12 | Not yet | AI isn't your next move. Write the process down and start keeping records. |
Notice what the bottom two bands don't say. They don't say "book a call anyway." If your score is under 20, nobody should be selling you a build this quarter, and a vendor who tries is telling you something useful about how they work.
What to do after you score
If you scored 26 to 30, the next question isn't whether to start. It's what to start with, and the answer isn't always the loudest problem. Read what to automate first before you pick, because sequencing is where most of the payback is won or lost.
If you scored 20 to 25, fix one thing and re-score. Write the process down, get the export working, or name the owner out loud. Most owners in this band are 30 days from ready, not six months.
If you scored under 20, do the unglamorous work first. Document the steps for two weeks as you do them, and start capturing the numbers you'd want a system to learn from. None of that is wasted, since any build would have needed it anyway.
Whichever band you land in, watch for a partner who plans, builds, and then supports the thing. A strategy deck with no build behind it leaves you where you started. That's the gap AI consulting for a small business should close, so ask any vendor how they handle month three.
Frequently Asked Questions
How long should an AI readiness assessment take for a small business?
About 20 minutes for one workflow. Enterprise assessments run for weeks and cost thousands because they cover departments, systems, and staffing across a whole organization. You're checking whether one process has a repeatable shape, findable data, and enough volume. That's a Tuesday afternoon question, not a quarter-long program.
Do I need clean data before I start?
Findable beats perfect. You need to name where the records live, get them into a spreadsheet, and have the last couple of years mostly complete. Messy formatting gets fixed during a build. If the honest answer is "it's all in my head," that's your first project, and it isn't an AI project.
What if I score well on everything except budget?
Then you're not ready this quarter, which is a legitimate answer rather than a failure. Either shrink the scope to the smallest slice that still saves real hours, or wait a quarter and go in fully funded. Don't start a build you can't afford to support, because support is where the value compounds.
Can I run this check without any technical help?
Yes. Nothing on the list requires knowing how a model works. Every dimension is about your process, records, volume, people, risk, and money, which you already know better than any consultant does. If you're unsure whether outside help is worth it, we wrote an honest guide to that decision.
How often should I redo this check?
Every six months, and any time the process or the software behind it changes. Businesses move. A workflow that scored 14 last spring can score 24 once you switch systems and start keeping records. Re-scoring takes 20 minutes, which beats assuming.
Run it this week
Pick your loudest workflow, score the six dimensions, and total it up. Twenty minutes gets you a defensible answer about whether AI belongs in your business right now, and that answer is worth more than another demo.
If you'd rather have the questions asked for you, the free 12-question AI Readiness Score scores it and hands you the result. It's self-serve and there's no call to book. If the honest answer is that AI wouldn't pay for itself in your business today, you'll have lost five minutes and gained a reason to stop shopping.
