AI Audit for Small Business: What It Checks and What You Get
An AI audit for a small business should rank where AI saves real hours and what it needs to work. What it checks, what it costs, and the findings to expect.
An AI audit for a small business should answer one question: which steps in how you actually work would AI take off someone's plate, and what would it need to do them? Done properly, it ends in a ranked list of three to five opportunities, each with the hours it saves, the data it depends on, and a rough cost. It takes anywhere from a 30-minute call to about three weeks, depending on how deep it goes.
That's not what most of the search results for "AI audit" describe. Most of them are about auditing an AI system you already run, for bias, security and regulatory compliance. Both are real services. They're bought by different people for different reasons, and it helps to know which one you're shopping for before anyone sends you a proposal.
Two services share the name "AI audit"
| Governance AI audit | AI opportunity audit | |
|---|---|---|
| The question it answers | Is the AI we already use safe, fair and compliant? | Where would AI save us real time, and is our business ready for it? |
| Who usually buys it | Companies with AI in production, often regulated or selling to enterprises | Owners and operators who haven't put AI into their workflows yet |
| What gets examined | Models, training data, outputs, access controls, documentation | Workflows, handoffs, the systems you run, and the state of your records |
| Measured against | Frameworks such as the NIST AI Risk Management Framework, ISO/IEC 42001, the EU AI Act | Hours spent per task, error rates, and what the data can support |
| Who should run it | Security, risk or compliance specialists | People who have built business systems and can tell you what's buildable |
| What you get | Findings against controls and a remediation list | A ranked roadmap of what to automate first, and what to leave alone |
A few small businesses need the governance kind. If you're in New York City and use software that screens job candidates, the city requires an independent bias audit of that tool every year. If an enterprise customer's security questionnaire asks how you govern AI, you may need one too. That work belongs with a compliance firm. We don't do it, and you shouldn't hire a systems builder to do it.
The rest of this post is about the opportunity audit, which is the one we run as the first step of our AI consulting work.
What a good opportunity audit actually checks
An audit that's worth anything looks at five things, in roughly this order. If a provider skips one, ask why.
- How the work really moves. Not the org chart and not the process document from 2022. Someone should walk through a real job or client file from first contact to final invoice and write down every place a person retypes, copies, chases or waits.
- Where the hours go. Each step gets a rough weekly time cost from the person who does it. Estimates are fine. What matters is that they come from the desk doing the work, not from the owner's memory of it.
- Whether the data exists. AI can only draft a quote from past quotes if past quotes are stored somewhere it can read, in a consistent shape. This is the step that kills more AI ideas than any other, and it's the one sales-led audits skip.
- What the risk is if it's wrong. A drafted follow-up email that's slightly off costs nothing. A wrong number on a payroll report costs a lot. Each opportunity needs a note on who checks the output before it goes out.
- What you already pay for. Plenty of software you own has AI features switched off or unused. An honest audit tells you when the fix is a setting, not a project.
A sample of what the findings should look like
The output that matters is a table, not a slide deck. Here's the shape we'd expect for a hypothetical 15-person HVAC service company with an office manager, a dispatcher and a part-time bookkeeper. Every number below is invented for illustration.
| Rank | Opportunity | Hours a week today | Data it needs | Data ready? | Effort |
|---|---|---|---|---|---|
| 1 | Turn technician job notes into a draft invoice | 8 | Job notes, price book, customer record | Yes, in the field app | Low |
| 2 | Draft maintenance-agreement renewal reminders | 3 | Agreement dates, equipment history | Partly; dates live in a spreadsheet | Low once the dates move |
| 3 | Answer "when's my tech arriving?" by text | 4 | Live schedule, technician location | Yes | Medium |
| 4 | Price replacement quotes from photos | 6 | Past quotes with equipment details | No; quotes are PDFs with no structure | High, after cleanup |
| Not now | AI phone receptionist | n/a | Call routing rules | n/a | Missed calls are already under 5%; low payoff |
Three things in that table are what separate a useful audit from a sales pitch. The ranking puts the dull, ready item first, even though photo quoting is the exciting one. The "Data ready?" column tells the owner that item 4 is a records problem before it's an AI problem. And the last row says no to something, with a reason. An audit that recommends everything hasn't done its job.
What an AI audit costs
We checked published prices on November 17, 2026. They vary because the word covers so much.
| Provider type | Published price | Source | What you typically get |
|---|---|---|---|
| Freelance marketplace listing | From about $250 | Upwork service listings | A written review of what you describe, usually without talking to staff |
| Small AI consultancy, entry audit | $500 to $1,500 | Bushe.co guide, May 2026 | Workflow mapping, a time-cost analysis and a ranked report |
| Consultancy assessment | $2,500 to $10,000 | AI Smart Ventures buyer's guide, 2026 | One to three weeks of work across several workflows |
| Free first call, fixed-fee assessment | Free, then quoted | Common among small firms, including us | A 30-minute triage, then a scoped assessment if there's real work |
The cheapest options aren't wrong for a very small business. The thing to watch is whether the audit includes time with the people who do the work. A report written from a questionnaire the owner filled in will describe how the owner thinks the business runs, which is usually not quite how it does.
The other trap runs the opposite way. A "free audit" that lasts several weeks has a cost somewhere, and it usually shows up in the price of the build that follows. Free should mean short.
Questions to put to anyone offering you an AI audit
- Will you talk to the people who do the work, or only to me?
- Will the report say which opportunities our data can't support yet?
- Do you resell any software? If so, which, and how are you paid for it?
- If the answer is "buy a per-seat AI subscription and stop there," will the report say that?
- Can a different developer build from your roadmap, or is it written for your team only?
- What will you recommend we leave alone?
The fourth question tells you the most. A provider who can't imagine recommending something cheap is selling you their next project.
Signs your business will get value from one
An opportunity audit earns its keep when work crosses more than one system or more than one person. The pattern we look for is the same information typed into two or three places: a job in the field app, then in the accounting package, then in a spreadsheet someone keeps for the owner. That's where AI inside your own systems does real work, because it can read one record and draft the next.
The published results we can point to came from exactly that kind of finding. At DS Water, replacing dozens of disconnected tools with one operating system took back more than 40 admin hours a week. At Fix My Books, automating client intake cut onboarding time by 80%. Neither started with AI as the goal. Both started with an honest map of where the hours went.
If you run a construction or trade business and want a quick self-check first, our AI operations scan for contractors takes a few minutes and points at the likely problem areas. And if the audit says to build, our AI implementation page covers how that part works.
When you can skip the audit
If you're a team of three working out of one tool, you probably don't need an audit. Buy a business plan for ChatGPT or Copilot, use it for a month, and see where it falls short. The same goes for a process that's still changing every few weeks: settle it first, because automating a moving target means paying twice.
How our AI audit works
Ours starts as a free 30-minute call with me or Carter. You walk us through how one job or client moves through the business, and we tell you on the call whether there's real work for AI and roughly where. Sometimes the answer is that a tool you already pay for covers it, and we'll say so.
If there's more to find, the next step is a fixed-fee assessment, quoted after that call, that takes one to two weeks. We map the workflow with the people who do it, check that the data exists, and hand you a ranked roadmap you own whether or not we build anything.
Bring one real example, a job, an order or a client file, and book a call with Ben. Thirty minutes with a real file will tell you more than any checklist.
Common questions
What is an AI audit?
The term covers two different jobs. A governance AI audit checks an AI system you already run for bias, security and compliance against frameworks like the NIST AI Risk Management Framework or ISO/IEC 42001. An AI opportunity audit looks at how work moves through your business and ranks the steps where AI would save time. Most small businesses that search the term want the second one.
How much does an AI audit cost for a small business?
Published prices we checked in November 2026 ran from a few hundred dollars for a freelancer's written review to $2,500 to $10,000 for a consultancy assessment. Many firms, including us, start with a free 30-minute call and quote a fixed fee for the full assessment after it.
How long does an AI audit take?
A first-pass audit call takes 30 to 60 minutes. A full assessment that maps workflows with the people who do the work and checks the data usually takes one to three weeks for a small business.
What should an AI audit report include?
A ranked list of opportunities with the hours each one saves, the data each one needs and whether that data exists today, a rough cost and effort for each, and the things it recommends you leave alone. If it only lists tools to buy, it's a sales document.
Do I need an AI audit before using ChatGPT?
No. If your team mainly wants help writing and researching, a ChatGPT or Microsoft Copilot business plan is a reasonable first step without an audit. An audit earns its cost when the work you want AI to take over runs through your own systems and records.
