AI audit for a small business: what it covers, what you get, and what it costs
An AI audit is a short, fixed-scope look at how a business runs that ends in a ranked list of where AI would pay and where it would not. Published prices for a small business run $1,500 to $5,000. This is what a real one examines, the document you should walk away with, and when you do not need one.
An AI audit for a small business is a one- to two-week review of the data, tools, workflows, people, and risk in the operation, ending in a ranked list of three to five opportunities with the hours each would save, what it would cost, and whether to do it now or not at all. Published prices run $1,500 to $5,000 for businesses under 50 people (ConsultKit). Ours takes two weeks at a fixed scope and ends in a plan that says what to start, what to defer, and what to leave alone.
What an AI audit is, and is not
An AI audit answers one question: where in this business would AI pay for itself, and where would it not. It is a review of the operation, not of the technology. The auditor looks at what the team does in a normal week, where the hours and fees go, which systems hold the data, and which decisions follow rules that could be written down. The result is a ranked list, and the ranking is the product.
It is not a strategy deck about the future of AI, and it is not a tool demo. A strategy deck tells you AI matters, which you knew. A demo tells you what one product does, which the vendor knew. Neither tells you which of your processes to hand off first, what that would cost, and what to leave alone. Lilach Bullock's guide sets the bar in one line: a proper audit should tell you at least one thing you were doing that AI cannot fix. We agree, and our version of that line is the "leave alone" column of the plan.
The reason the audit matters is in the failure research. RAND found more than 80 percent of AI projects fail, and the first cause it names is that the business and the builders never agreed on the problem. An audit is the agreement, on paper, before anyone builds.
What it examines
Published descriptions of the work converge on five areas, and ours covers the same ground with different inputs.
- Data. Where customer, sales, and operations information lives, and whether a system could reach it. We gather financial records and system activity for this.
- Tools and systems. What software is in use, what it already does that nobody turned on, and what it cannot do. We read the process documents and the systems themselves.
- Workflows. The step-by-step sequence of the work that eats the most hours, mapped plainly enough that a stranger could follow it.
- People. Who owns each outcome, what knowledge lives in one head, and whether the team has the time and appetite to adopt a change. We collect this as team knowledge, in conversation.
- Risk. What data may leave the building, what the system must never decide alone, and which workflows are regulated.
Every opportunity that comes out of those five is weighed the same way: operating value, feasibility, risk, and ownership. An opportunity with high value and no owner ranks below one with moderate value and a named person who wants it.
The document you should receive
The deliverable is a short document, and the test of a good one is that you can act on it the day after you read it. Bullock describes it as the top three to five opportunities, each with an estimate of hours saved per week, a rough cost to implement, and a plain answer on whether it is worth doing this quarter or next year. That matches what we deliver.
Our version is a ranked plan with three columns: what to start, what to defer, and what to leave alone. The first column has one item at the top with a first move attached. The third column is the one most audits skip, and it is where the honesty lives: the processes that looked like AI candidates and are not, because the work changes shape every time, or nobody owns the result, or the rule is still an opinion.
If the document you receive has a technology on every line and no process, you bought a vendor pitch with a cover page.
What an AI audit costs
Two published guides put numbers on it, and they agree within the band.
| Business size | Price | Effort | Source |
|---|---|---|---|
| Under 10 employees | $1,500 to $3,000 | 4 to 8 hours | ConsultKit |
| 10 to 50 employees | $2,500 to $5,000 | 8 to 12 hours, three to five core processes | ConsultKit |
| 50 to 500 employees | $5,000 to $15,000 | 15 to 30 hours, full department mapping | ConsultKit |
| 1 to 5 people, one core process, UK | £1,500 | One to two weeks | Lilach Bullock |
| 15 to 30 staff, several departments, UK | £6,000 and up | One to two weeks | Lilach Bullock |
| Discovery engagement, written plan | $2,000 to $5,000 | Not stated | Layer3 Labs |
Prices as published by ConsultKit (March 22, 2026), Lilach Bullock (2026), and Layer3 Labs (updated June 17, 2026). ConsultKit adds 30 to 40 percent for deep industry expertise.
We do not publish one flat price for our AI Audit, because a three-person firm and a forty-person firm are not the same two weeks. The scope is fixed and the price is quoted before work starts, in writing, with the schedule. Held against the table, it lands inside the small-business band.
How ours runs over two weeks
Week one is gathering. We collect financial records, process documents, team knowledge, and system activity, and we watch the work move. Week two is weighing. Every candidate process is scored on operating value, feasibility, risk, and ownership, and the scores become the ranking.
The audit ends in the ranked plan and a first move. The first move is usually one of two things: an existing tool that covers the top process, which the team can be using within the month, or a scoped custom build when the work is hard in the gap between two systems that no product connects. Our dental payment integration came out of exactly that second kind of finding.
Two minutes before two weeks.
The ten questions run a first pass on the same seven dimensions an audit examines. If one process is already obvious, you may not need the audit at all, and the result will say so.
When you do not need one
An audit is for a business that cannot yet name its first process, or that has several candidates and no way to rank them. If you can already point to one task that repeats every week, arrives in the same shape, has one owner, and costs you a number you can state, skip the audit. Go to the tool step. Our AI Finder shortlists existing products for free, and Buy Your Time Back does the whole tool step for $995, refunded if the team does not get five hours back in a normal week.
An audit is also the wrong purchase when the business is not defined enough to audit. If the work changes shape every time and nobody owns the outcome, the two weeks will produce a document that says so, and you can produce that yourself. The free readiness assessment is built for that reading.
Four questions before you buy one
- What does the deliverable look like? Ask to see a redacted one. It should be short, ranked, and have a column for what not to do.
- Who does the looking? The person who reads your process documents and talks to your team should be the person who writes the plan.
- Is the auditor selling the build? Not disqualifying, since we do both, but the audit must be able to end in "buy a tool" or "not yet" without penalty to the auditor.
- What is the price tied to? A fixed scope and a schedule, in writing, before work starts.
An audit that passes those four costs less than one month of the process it ranks first. That is the arithmetic that makes it worth buying.
What to hold on to
- An audit ranks processes, not technologies. Five areas examined, every candidate weighed on value, feasibility, risk, and ownership.
- The deliverable is short and ranked. Three to five opportunities with hours, cost, and a verdict. A "leave alone" column is the mark of an honest one.
- Published prices sit in the low thousands for small businesses. Effort of 4 to 12 hours over one to two weeks; larger firms run higher.
- Ours is two weeks, fixed scope, quoted in writing. Ends in a plan that says what to start, what to defer, what to leave alone.
- Skip it if you can already name the process. Go to the tool step instead; the free ten questions tell you which reading you are.
Questions owners ask us
What is an AI audit for a small business?
A short, fixed-scope review of how the business runs: its data, tools, workflows, people, and risk. It ends in a ranked list of where AI would pay for itself and where it would not, with the hours each opportunity would save, what it would cost, and whether to do it now. It is a review of the operation, not of the technology.
How much does an AI audit cost for a small business?
Published 2026 prices run $1,500 to $3,000 for businesses under ten people and $2,500 to $5,000 for ten to fifty, with larger firms at $5,000 to $15,000 (ConsultKit). UK guides quote £1,500 to £6,000 and up (Lilach Bullock). Our AI Audit is two weeks at a fixed scope, priced per business and quoted in writing before work starts.
How long does an AI audit take?
One to two weeks for a business under 15 people looking at two or three core processes, according to published guides. Ours takes two weeks: one to gather records, documents, team knowledge, and system activity, and one to weigh and rank every candidate process.
What do you get from an AI audit?
A ranked plan. Ours has three columns: what to start, with a first move attached; what to defer; and what to leave alone, because the work is not defined enough for a system. Each opportunity carries the hours it would save, the rough cost, and who would own it.
Do I need an AI audit before using AI?
No. If you can already name one weekly process, price what it costs today, and check whether an existing tool covers it, go straight to the tool step. The audit is for businesses that cannot yet name the first process or have several candidates and no way to rank them.
What is the difference between an AI audit and an AI readiness assessment?
A readiness assessment is a quick self-check, usually free, that tells you whether there is a process to hand off and a reason to do it now. Ours takes about two minutes. An audit is paid, takes one to two weeks, examines the operation in detail, and ranks the specific opportunities with costs attached. Do the assessment first; buy the audit if the reading calls for it.
Sources
Prices and scope descriptions are quoted from published guides; our own terms are quoted from the services page. Checked September 5, 2026.
- Lilach Bullock, "AI Audit for Small Business: What It Includes (And What It Doesn't)". 2026. The five areas, the deliverable, the one-to-two-week duration, and UK pricing.
- ConsultKit, "How to Price an AI Audit for a Small Business (With Real Numbers)". March 22, 2026. Price tiers by business size and effort.
- Layer3 Labs, "AI Consulting Rates and Pricing in 2026". Updated June 17, 2026. The discovery engagement band.
- RAND Corporation, "The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed". The 80 percent failure rate and misalignment on the problem as the first cause.
- AI Audit, our two-week engagement as described on the services page, and the dental payment integration that came out of one.