AI Readiness Assessment for Small Business: Score Your Records First
An AI readiness assessment for a small business comes down to your records. Score five of them in ten minutes, see what each score means, and what to fix first.
An AI readiness assessment for a small business comes down to one thing more than any other: the records the AI would have to read. Strategy, culture and infrastructure matter in a 5,000-person company. In a 20-person one, the question is whether your customer list, job history, pricing and documents live somewhere consistent enough for software to use. The scorecard below takes about ten minutes and tells you which of your records is ready and which one is holding you back.
I'm Carter. I run delivery on our AI consulting work, which means I'm the person who finds out on day one of a build whether a business was actually ready. The answer is decided by the records far more often than by budget or enthusiasm.
Why the enterprise checklists don't fit a 20-person company
Most of what you'll find under this search was written for large organizations. Cisco's AI Readiness Index, the best-known one, surveyed 8,000 senior IT and business leaders across 30 markets in 2025 and found that about 13% of companies were fully ready for AI. Its pillars, and the ones most consultancy frameworks copy, look like this. Here's what each one turns into when the company is small.
| Enterprise pillar | What it asks at scale | What it really means at 10 to 50 people |
|---|---|---|
| Strategy | Is there a board-approved AI roadmap? | Can you name the one job you want off someone's desk? |
| Infrastructure | Are networks and compute ready for AI workloads? | Almost never the problem. You'll rent the AI from a provider. |
| Data | Is enterprise data governed, catalogued and accessible? | The whole question. Scored below. |
| Governance | Are there policies for risk, bias and compliance? | Who checks the AI's output before a customer sees it? |
| Talent | Do we have data scientists and AI engineers? | Is there one person who'll own the system once it's built? |
| Culture | Will the workforce adopt AI? | Will the person doing the task today use the draft it gives them? |
Four of those six rows become a single sentence each. That's not a knock on the frameworks. They're answering a different company's question. A small business doesn't need a readiness program. It needs to know whether the data behind one job is usable, and that's what the rest of this page measures.
If you're still deciding whether to bring anyone in at all, our self-check on whether you need an AI consultant scores the size of the problem. This page scores the foundations under it. They're different tests, and you can pass one and fail the other.
Score five records in ten minutes
The AI jobs we build for service businesses read from some mix of five records. Score each one from 0 to 3 using the descriptions in its row. Score honestly against the way things are today, not how they'll be once someone gets around to the cleanup.
| Record | 0 | 1 | 2 | 3 |
|---|---|---|---|---|
| Customers and contacts | In inboxes and phones | A spreadsheet, plus duplicates elsewhere | One CRM or app, some gaps | One system, every customer, no duplicates |
| Jobs, orders or matters | Paper or whiteboard | Calendar entries and notes | An app, but notes are free text | An app with status, dates and assignee on every record |
| Pricing and past quotes | In the owner's head | Old PDFs or Word files | A price list, quotes not linked to jobs | Price book plus quotes stored against each job |
| Documents and forms | Paper | Scans in shared folders, no naming rule | Digital, attached to the right record most of the time | Digital, attached to the record, same template each time |
| Customer messages | Personal cell phones | One shared inbox, nothing logged | Logged for some customers | Calls, texts and email logged against the customer |
Then answer two yes-or-no questions about the record that matters most for the job you have in mind:
- Does it go back at least twelve months? AI that drafts a quote needs past quotes to learn your pricing habits. A system you started in September doesn't have them yet.
- Can software reach it? That means an export, an API or an integration your current tool already supports. A record locked inside an app that won't let it out is a 0, whatever its quality.
Reading your total
| Total (out of 15) | Where you stand | What to do next |
|---|---|---|
| 0 to 5 | Not ready for AI on your own data | Use a ChatGPT or Copilot business plan for writing and research. Put the money toward getting one record into one system. |
| 6 to 10 | Ready for one narrow pilot | Pick the job whose records scored 2 or 3, and pilot there. Leave the low-scoring records alone for now. |
| 11 to 15 | Ready to build | The limit is no longer your data. It's choosing the right first job, which is what an audit is for. |
The total matters less than the lowest score on the record your chosen job depends on. A business can score 12 and still not be ready for AI quoting, if pricing and past quotes is the row that scored 0. Look at that row before you look at the sum.
A worked example, scored
Say you run a hypothetical 20-person plumbing and drain company. The numbers here are invented to show how the scoring works. The owner wants AI to draft repair quotes from a technician's photos and notes, because quoting eats the evenings.
They score it like this. Customers: 2, the field app has them but there are duplicates from the old spreadsheet. Jobs: 2, every job is in the app but the notes are free text. Pricing and past quotes: 1, there's a price list, and quotes go out as PDFs from a template that isn't linked to the job. Documents: 2. Messages: 1, everything goes through one office inbox and nothing is logged. Total: 8 out of 15.
By the total, that's pilot territory. By the row that matters, it isn't. Photo quoting depends on past quotes tied to the work that was done, and this business stores neither the link nor the line items. So the honest plan is two steps. First, start saving quotes as structured records against each job, which takes a few weeks of normal work to build up. Then, once there are a few months of them, the AI has something to learn from.
In the meantime, jobs scored a 2. A smaller pilot that reads finished job notes and drafts the invoice could start straight away, and it would pay for some of the quoting work while the data builds.
That's roughly the shape of what we found at Grit Construction. Estimates, change orders, invoicing and job costs had to live in one system before anything could be automated, and putting them there took 10+ hours a week of invoicing and reminders off the owner's plate. The next step we've published for Grit, pricing a change order against the original estimate, only makes sense because those records now sit together.
Free online assessments are worth ten minutes
There's no shortage of free readiness quizzes. Microsoft publishes one on Microsoft Learn, and most AI consultancies have their own lead-capture version. They're a fine way to start thinking. Two things to know before you trust the result:
- Most are written for bigger companies. You'll be asked about AI policy, executive sponsorship and model governance. A low score there doesn't mean much for a business whose real gap is a quoting spreadsheet.
- Most end in a sales call. That's not a scam, it's how free tools pay for themselves. Just read the result as a conversation starter, not a diagnosis.
If you're in construction or the trades, our AI operations scan for contractors is the free version we built for that audience. It asks about the work itself rather than about AI strategy.
When a paid assessment beats a self-score
The scorecard above works when the records you care about are visible to you. A paid assessment earns its fee when they aren't: when the data is spread across four apps and nobody can say which one is right, when a former employee set up the system and nobody has the admin login, or when the job you want automated crosses three people and you only see one part of it.
It's also worth paying for when the decision is expensive. If you're about to sign an annual contract for an AI add-on, a few hours of someone checking whether your data can feed it is cheap insurance.
We don't sell readiness as a separate product. It's the first half of the AI audit we run at the start of our AI consulting work: we check the records first, because there's no point ranking opportunities your data can't support. The audit post covers what that costs and what the findings look like. If the answer is to build, our AI implementation page covers how that part runs.
Fix the lowest row before you buy anything
If you scored yourself and one row came out at 0 or 1, you already know your first project, and it isn't an AI project. Getting quotes linked to jobs, or messages logged against customers, is ordinary systems work. It's also the groundwork behind every result in our published case studies. At DS Water, pulling dozens of disconnected tools into one operating system took back more than 40 admin hours a week before any AI went in.
If you'd like a second opinion on your score, bring the scorecard and one real job file and book a call with Ben. Thirty minutes is usually enough to tell you which row to fix and whether AI belongs on the list this year.
Common questions
What is an AI readiness assessment?
It's a check on whether your business can put AI to work on a specific job and get a reliable result. For a small business, the deciding factor is almost always the state of the records the AI would read: where they live, how consistent they are, and whether software can reach them.
How do I know if my business is ready for AI?
Pick the task you'd most like AI to take over and look at the records behind it. If they live in one system, use the same fields every time, go back at least a year, and can be exported or reached through an API, you're ready to pilot. If any of those fail, fixing that comes first.
Is there a free AI readiness assessment?
Yes. Microsoft publishes a free AI Readiness Assessment on Microsoft Learn, and many consultancies offer free online quizzes. They're useful for a first read. Most are written for larger organizations, so they spend more time on strategy, governance and talent than on the records a 20-person business actually depends on.
What is the difference between an AI readiness assessment and an AI audit?
A readiness assessment asks whether your business can support AI on a given job. An AI opportunity audit goes further and ranks which jobs to automate first, with the hours each one saves. Many firms, including us, run the readiness check as part of the audit rather than selling it separately.
What are the pillars of AI readiness?
Enterprise frameworks usually name five to seven: strategy, data, infrastructure, governance, talent and culture. In a small business most of those collapse into two questions: are the records good enough, and is there one person who will own the result.
