Before You Hire an AI Consultant, Run This 30-Minute Audit
Nearly every founder I work with starts with the same question: “How do we use AI?” It sounds reasonable. In my experience, it’s also the costliest place to begin.
The order of the question matters. When you start with “How do we use AI?” you put the tool in the driver’s seat and your business in the passenger seat. Flipping it to “Where would AI actually help?” puts your business first and the technology second. Same words, different outcomes. One leads to a proposal. The other leads to a real decision.
I’ve seen too many consultants jump straight to the proposal, because that’s what gets billed. A 30-minute audit, run by you, not a vendor, will tell you if you actually need outside help. The teams who do spot it quickly. The ones who don’t save themselves a meeting and twelve weeks of distraction. Eight questions, three sections. Make sense of your answers before you book the call.
Eight questions, three sections, one timer.
I designed this 30-minute audit as an eight-question self-diagnostic. It tells you whether AI belongs on next quarter’s roadmap, and whether bringing in a consultant is the right move. Eight questions, three sections. Set a timer. If you’re still working at 31 minutes, the audit is already surfacing the real work.
The sections run in a deliberate order. Work first, evidence second, judgement last. Most AI conversations skip work and jump to tools. They skip evidence and jump to outputs. They never touch judgement at all. The order is the point.
Where the work happens — three questions.
The first three questions focus on where the work happens. Most AI conversations skip this and jump straight to tools. Mapping the shape of your week is always the better starting point.
- Which three tasks in your business cost the most time each week?
- Of those three, which involve reading, summarising, categorising, or drafting? These are AI’s strongest hands.
- Who would lose the most if one of those tasks was done badly?
No baseline, no signal — three questions.
The next three questions dig into evidence. AI buying decisions made without a baseline produce stories, not signal. These questions force you to set the baseline before any rollout.
- Can you point to a number — hours saved, errors caught, revenue attributed — that would prove AI worked in your business?
- Do you currently have a way to measure that number without AI? If no, you do not have a baseline. You have a hope.
- What would you commit to a three-month trial: budget, time, and attention?
The line you won’t cross — two questions.
The last two questions ask you to define the line you’re not willing to cross. Most teams discover that line only after they’ve already crossed it.
- Name one decision in your business you would never delegate to software. Why not?
- If the AI got it wrong on a Monday, how would you know by Friday?
If you can answer all eight in 30 minutes, you’re ahead of most businesses I’ve worked with. If you can’t, that’s your audit result. The gap between “we should use AI” and a clear answer to question seven is the difference between a working pilot and an expensive habit.
Five patterns cover most of what the audit surfaces.
The pattern of your answers matters more than the answers themselves. Five patterns cover most of what comes up.
Pattern 1 · Specific tasks, real baseline.
You’re ready to run a pilot. Bring in an AI consultant to design it well, or run it yourself with a clear hypothesis, a measurement plan, and an exit criterion. The discipline behind a first AI project that’s worth shipping isn’t about model choice. It’s about the planning. The model can change. The planning can’t.
Pattern 2 · Tasks but no baseline.
This is an AI-readiness problem, not an AI problem. Fix the measurement layer first. Without it, every result is a guess in expensive packaging. I tell teams to spend a week on the baseline. That way, the pilot becomes a real test, not a press release.
Pattern 3 · Baseline but the tasks are all judgement calls.
AI isn’t your highest-leverage investment right now. Solve the judgement problem with people, process, or better tooling. Come back to AI when there’s repeatable work for it to do. AI is at its best where the same kind of thinking happens over and over. If every task is one-of-a-kind, you’re buying a hammer for a job that needs a steady hand.
Pattern 4 · Could not answer question seven.
Stop here. Any rollout that starts without an answer to that question builds a system you can’t supervise. This isn’t academic. It defines the line between AI as a tool and AI as a vendor running parts of your business. Write the answer down before the next vendor call. That line is your answer.
Pattern 5 · Could not answer question eight.
You don’t have feedback loops yet. Build those before you build AI on top of them. AI without feedback loops becomes a confidence machine — it keeps telling you it’s working long after it’s stopped. This is often the audit’s quietest but most important finding.
Hire when getting it wrong costs more than the engagement.
Hire when you’re in pattern 1 or 3 and don’t have the internal capacity to run the pilot. Hire when you need an outside voice to align a leadership team pulling in different directions on AI. Hire when the cost of getting the rollout wrong is bigger than the cost of the engagement. The good engagements pay for themselves in mistakes avoided, not features added.
Wait when you’re in pattern 2 or 4. Wait if nobody in the business has named a metric that matters. Wait if the brief is “let’s try some AI stuff” instead of a specific problem with a clear outcome. Waiting can feel expensive. Hiring without a brief is always more expensive.
My view, based on the businesses I’ve spoken to over the past year: most that contact a consultant too early waste the engagement. Most that wait too long miss the window. The audit is what moves you from too early to ready. If your eight answers say wait, then wait. The same consultant will still be there in three months, and you’ll arrive with a brief worth answering. Three months of clarity beats three months of unfocused momentum.
Three questions separate the consultants who deliver from the ones who sell.
- Show me a pilot you ran that did not work, and what you learned. If they can’t answer, they haven’t run real pilots. Every honest practitioner has a null result somewhere. The story of what they changed afterwards is the story of how they think.
- What baseline do you need from me before you start? If the answer is “we’ll figure that out,” run. Baseline first is the difference between evidence over instinct and a confident invoice. The consultant who measures before they build is the one who can prove what changed.
- When would you tell me to stop the project? Consultants with no exit criteria are selling optimism. The good ones write the stop conditions into the proposal. They know the second-best answer is a clean kill, and they will not bury that option to keep the engagement going.
If you’ve run the audit and want someone to interpret the answers with you, you can book an AI Opportunity Roadmap call here.
Frequently asked questions
00 entries-
Q.01
What is the 30-minute AI consultant audit?
A self-diagnostic of eight questions across three sections — work and workflows, evidence and baseline, judgement and risk. Run by you, not a vendor. The aim is to decide whether you actually need outside help before you book the call. The teams who need a consultant spot it quickly. The ones who don’t save themselves a meeting and twelve weeks of distraction.
-
Q.02
Should I hire an AI consultant or wait?
Hire when the cost of getting the rollout wrong is bigger than the cost of the engagement, when you don’t have the internal capacity to run a pilot, or when leadership is pulling in different directions on AI. Wait when nobody in the business has named a metric that matters, when the brief is “let’s try some AI stuff” rather than a specific outcome, or when the audit surfaces a missing baseline or a missing feedback loop. Most who contact a consultant too early waste the engagement.
-
Q.03
What are the eight questions in the AI consultant audit?
Eight questions across three sections. Work and workflows: which tasks cost most time, which involve reading/summarising/categorising/drafting, who loses most if a task is done badly. Evidence and baseline: can you point to a number that would prove AI worked, can you measure it without AI, what would you commit to a three-month trial. Judgement and risk: one decision you’d never delegate to software, how you’d know on Friday if AI got it wrong on Monday.
-
Q.04
Why does the order of the question matter — “How do we use AI?” versus “Where would AI actually help?”
When you start with “How do we use AI?”, you put the tool in the driver’s seat and your business in the passenger seat. Flipping it to “Where would AI actually help?” puts your business first and the technology second. Same words, different outcomes. One leads to a proposal. The other leads to a real decision.
-
Q.05
What does it mean if I can’t answer questions seven or eight?
Question seven asks you to name a decision you’d never delegate to software. If you can’t answer it, any rollout you start builds a system you can’t supervise — write the answer down before the next vendor call. Question eight asks how you’d know on Friday if AI got it wrong on Monday. If you can’t answer it, you don’t have feedback loops yet. Build those before you build AI on top of them. AI without feedback loops becomes a confidence machine.
-
Q.06
What three questions should I ask any AI consultant before hiring them?
Show me a pilot you ran that did not work, and what you learned. What baseline do you need from me before you start? When would you tell me to stop the project? If they can’t show a null result, they haven’t run real pilots. If they can’t name the baseline they need, run. If they don’t have exit criteria, they’re selling optimism.
-
Q.07
When is it too early to hire an AI consultant?
When you’re in Pattern 2 (tasks but no baseline) or Pattern 4 (couldn’t answer the supervision question). Wait if nobody in the business has named a metric that matters or if the brief is “let’s try some AI stuff” rather than a specific outcome. Waiting can feel expensive. Hiring without a brief is more expensive. The same consultant will still be there in three months, and you’ll arrive with a brief worth answering.
-
Q.08
What’s the difference between an AI-readiness problem and an AI problem?
An AI problem is when you have a specific task with a clear baseline and you’re choosing how to use AI on it. An AI-readiness problem is when you don’t have the measurement layer yet. Fix the measurement first. Without it, every AI result is a guess in expensive packaging. Spend a week on the baseline — that way the pilot becomes a real test, not a press release.
