If you are thinking about an AI audit for your business, ask what you will have in your hands when it is done. A good audit ends with a written list of specific fixes, ranked, with an owner for each. A weak one ends with only a score and a slide deck.
This post explains what a useful audit for a service business covers, what to check, and how to spot one that will not help you.
Two Things Called an AI Audit, and Which One This Is
The phrase covers two different jobs, and search results mix them together.
A compliance or model audit tests an AI system for bias, accuracy or legal fit. It is aimed at companies that build or sell AI products, and usually involves lawyers and technical reviewers.
A workflow audit looks at how your business runs today and where AI could remove delay, errors or lost revenue. It is aimed at companies that want to use AI, and it looks at your leads, follow-up, reporting, tools and people.
This article is about the second kind. If you run a roofing company, a home service business or a professional services office, that is almost certainly the one you want.
What Gets Mapped: Tools, Lead Path, Handoffs, Reporting
The first phase is a map of how work moves through your company today. It should cover:
- Tools. Every system your team uses, including the ones nobody mentioned: the shared inbox, the spreadsheet someone keeps, the text line on a personal phone.
- The lead path. Where leads come from, who sees them first, how fast they get a reply, and what happens when they do not answer.
- Handoffs. Each point where work passes from one person or system to another. These are where things get dropped.
- Reporting. What numbers you look at, who builds them, and whether anyone trusts them.
Our post on where a business leaks revenue lists the common gaps this phase tends to find, so read it first if you want a preview.
The output of this phase is a plain diagram or list you could hand to a new hire. If the auditor cannot describe your current process back to you accurately, the later findings will not be reliable.
What Gets Checked: Data Quality, Access, and Who Owns Each Account
Next comes a check of what the tools actually hold and who can reach them.
- Data quality. Are contact records complete and current? Are duplicates piling up? Are lead sources recorded or left blank? AI built on messy records produces messy output.
- Access. Who has logins to each system? Are there former employees or past vendors who still do? Are people sharing one password?
- Ownership. Is each account in your company’s name or in an employee’s or contractor’s name? If the person who set up your phone system left tomorrow, could you still get in?
Account ownership is the item owners most often wish they had checked sooner. You should leave the audit knowing, for every system, who holds the keys.
The Team Layer: Habits, Shadow Tool Use, Training Gaps
Tools are half the picture. The other half is what your people do.
An audit should ask, in conversation and not in a survey that nobody answers:
- Which AI tools are staff already using, on their own, for work?
- What are they putting into them? Customer names, pricing, contract text?
- Who is confident with these tools, and who avoids them?
- Where do people say they lose time?
Shadow tool use is common. A good audit treats it as information and not as a reason to scold anyone. It tells you what your team wishes the company provided.
The training gaps found here feed directly into a plan for AI training if you decide to run one.
Risk Review: Customer-Facing Output, Private Data, Approvals
This section looks at what could go wrong and how you would catch it.
Customer-facing output. Anything a customer reads, from a text to a proposal, should be approved by a named person before it goes out. The audit checks whether that is true today, including for the automations you already have.
Private data. Which categories of information are allowed into which tools? Customer financial details, health information and signed contracts need clear rules.
Approvals. For each automated step, who can pause it, who reviews it, and who is told when something fails?
AI can draft quickly. It can also be confidently wrong. The safeguard is a person who approves, and an audit should show you where that person is missing.
The Deliverable: A Ranked List of Fixes with Owner and Effort
The written findings are the product. Ask to see a sample before you commit. A useful one has:
- A ranked list of fixes. Highest value first.
- For each fix, what is wrong and where it shows up. Specific enough that someone could go and look.
- An owner. A person on your team, or the builder, responsible for the fix.
- An effort level. Small, medium or large, in plain terms, so you can pick quick wins first.
- What to leave alone. Things that are working and should not be touched.
- A recommended first step. One workflow to start with, and how you will measure it.
You should be able to read it in one sitting and know what to do on Monday. If you are a contractor deciding where to begin, where a contractor should start with AI shows the one-workflow test an audit should feed.
Red Flags to Watch For
Be cautious of an audit that:
- Ends in a number and nothing else. “Your AI readiness is 62,” with no ranked fixes attached. A score is useful for checking progress later, but on its own it tells you nothing to do next.
- Is the same for every client. Generic findings mean nobody looked at your business.
- Needs access to everything on day one. A good audit asks for what it needs and says why.
- Recommends a long list of new software. Findings should start with what you already own.
- Cannot name who does the work after. An audit with no path to a fix is a report that gets filed.
- Promises specific results. Nobody can promise revenue before they have seen your numbers.
Questions to Ask Before You Hire Anyone to Audit
Put these to any provider:
- Can I see a sample of the written findings, with the client details removed?
- Who will actually talk to my staff, and how long will it take them?
- What do you need access to, and will it be read-only?
- Which findings would you expect in a business like mine?
- What happens after the findings: who does the work, and what does it cost?
- If the findings say I need very little, will you tell me so?
The last question tells you the most. An auditor who cannot say “you are mostly fine, fix these two things” is selling the next project.
How the AI Leak Audit Works, and What It Costs You to Start
Our free offer is the AI Leak Audit, a 20-minute teardown of how leads, follow-up, handoffs and reporting run in your business today. You get a map of where leads, time and revenue slip between your tools, the one fix that would move your numbers most, and an honest estimate of what it takes to build. You keep it either way.
If the findings point to a bigger project, that is where AI consulting or building an AI operating system for your business comes in. If they do not, you have a clear list you can hand to your own team.
It costs you a short conversation to start and the time to answer questions about how your business runs.