Here's the short answer: ask about process before tools, proof before promises, and handover before lock-in. A good AI consultant or coach can walk you through a real result with real numbers, will tell you plainly what AI won't fix, and has a plan for making themselves unnecessary. The ten questions below sort those people from the rest — before you've spent a dollar.
The problem is rarely the tech
Enterprise AI research is useful context, but it does not predict whether your engagement will succeed. When hiring, ask who will own the process, how it fits your work, and how you will measure a trial. Read the research context and its limitations before treating a headline statistic as a sales promise.
The wrong helper doesn't just waste your fee. They waste your data, your team's goodwill and six months of momentum. So before you compare anyone's prices, compare their answers.
The 10 questions to ask
Ask them in order if you like — the first three do most of the filtering. You don't need to be technical to ask any of them, because every one is about how the person works, not how the models do.
1. "Walk me through a real result, with numbers."
Anyone can say they've "helped businesses transform with AI". Numbers force specifics: what was the task, how many hours did it take before, what does it take now, and how was that measured?
A good answer sounds like: a specific story with a baseline and an after — "their weekly reporting took two days; we rebuilt the process and automated it, and now it's a couple of hours" — including something that went wrong along the way. Real work always has a wrong turn in it.
A bad answer sounds like: logos, adjectives and percentages with no source. "We've delivered incredible efficiencies for dozens of clients" is not a result. It's a slogan.
2. "What won't AI fix in my business?"
This is the honesty test. AI is genuinely good at repetitive, rule-based work and genuinely bad at judgement calls, relationships and broken processes. Someone who knows the edges will name them without flinching.
A good answer sounds like: a real list — "it won't fix an unclear offer, it can't make judgement calls you haven't defined, and if your process is a mess we fix that first, or we're just automating the mess."
A bad answer sounds like: "with the right platform, almost anything is possible." If everything in your business looks automatable to them, they're not looking at your business — they're looking at your budget.
3. "What happens when you leave?"
This one exposes lock-in. Some providers build everything inside their own accounts and platforms, so ending the engagement means losing the system. You want the opposite: an exit that's planned from day one.
A good answer sounds like: "everything runs in your accounts, you get the documentation, and I train your people to run and change it. My job is to become unnecessary."
A bad answer sounds like: "most clients keep us on a monthly retainer to manage it." Ongoing help should be your choice, not a hostage situation.
4. "Who actually does the work?"
A classic agency move is selling you the impressive founder and delivering the work through juniors you never meet. That isn't automatically bad — but you should know before you sign, not after.
A good answer sounds like: a straight answer with names — "me, personally", or "I scope it, my developer builds it, and you'll meet them in week one."
A bad answer sounds like: "our delivery team takes care of that" — no names, no seniority, and nobody you can actually talk to.
5. "How do you handle our data?"
The fastest way to turn an AI project into a liability is pasting customer data into free consumer tools. Whoever you hire will be touching your information, so they should have thought about this harder than you have.
A good answer sounds like: specifics — business-tier tools where your data isn't used for training, clarity about exactly what goes where, and access that gets revoked when the engagement ends.
A bad answer sounds like: "don't worry, it's all secure." If they can't name which tools your data passes through, they don't know.
6. "What's the total price?"
Transparent pricing is one of the strongest green flags in this industry, because it signals confidence and respect. Vague pricing usually means the price depends on how much they think you'll pay.
A good answer sounds like: a number, a rate card or a public pricing page — plus what's not included. If you want market benchmarks first, I've broken down what AI coaching costs in Australia separately.
A bad answer sounds like: "let's start with a discovery engagement and scope from there" — where the discovery costs thousands and the real quote only arrives once you're already invested.
7. "What do we do in week one?"
Doers and deck-makers answer this question very differently. You're looking for someone who touches real work in week one, not someone who schedules six weeks of workshops first.
A good answer sounds like: "we pick one repetitive task, I watch how it actually happens today, and we get a first version working." Small, concrete, real.
A bad answer sounds like: "we begin with an AI-readiness assessment and a strategic roadmap." A roadmap that arrives before anything runs is a very expensive PDF.
8. "Which tools would you recommend — and why those?"
Plenty of "consultants" are effectively resellers, with a commission on whatever they recommend. The tool should be chosen for your problem, not for their margin.
A good answer sounds like: "it depends what you already use — often the tools you have will do the job." And any partnerships or referral fees get disclosed unprompted.
A bad answer sounds like: the same branded platform for every client and every problem — especially if they happen to sell licences for it.
9. "What would make you tell us NOT to do this?"
An advisor with no walk-away conditions isn't advising you — they're billing you. This question tests whether "no" is anywhere in their vocabulary.
A good answer sounds like: real conditions — "if the process is too broken, if your data isn't ready, if a simple spreadsheet fix solves it, or if nobody on your team has time to own it, I'll tell you to wait."
A bad answer sounds like: a pause, then a pivot: "honestly, in this market, no business can afford to wait on AI." That's a sales script, not advice.
10. "How will we measure success?"
Define how you will judge the trial. If nobody defines success before the work starts, nobody can prove it happened — and "it feels faster" won't survive a bad quarter.
A good answer sounds like: "we take a baseline first — hours per week on the task, error rate, turnaround time — then measure the same things after, and review on a date we set now."
A bad answer sounds like: "adoption", "engagement", "AI maturity" — activity metrics that sound like progress but can't be tied to a saved hour or a saved dollar.
Notice the pattern: not one of the ten questions is about the technology. They're about proof, honesty, process and handover — the four things that actually decide whether you get a result.
Red flags: end the conversation
- Guaranteed ROI. "Guaranteed 5× return" isn't confidence — it's bait. Honest operators guarantee their effort and their process, never someone else's outcome.
- Tool-pushing. One platform recommended for every problem, especially when they resell it or clip a commission on the licence.
- Jargon walls. If they can't explain it to you in plain English, they can't teach it to your team. Complexity is often camouflage.
- No curiosity about your actual work. A proposal that arrives before they've asked how the job really gets done is a template with your logo on it.
- Hidden pricing. A price that only appears after a paid "discovery engagement" is a price designed around your sunk costs.
Green flags: lean in
- Verified before-and-after numbers they can walk you through step by step — including what went wrong.
- Plain English. They make AI feel smaller and more doable, not bigger and scarier.
- They start small. One task, one process, weeks not quarters — something in production before anything on a roadmap.
- They teach you to run it yourself. Documentation, training and your own accounts — handover by design, not dependence by design.
- They'll say "AI won't fix that." The cheapest advice you'll ever get is the project you didn't do.
Where I fit — the honest version
Tell me what you want to achieve. I’ll recommend a suitable approach and provide a clear scope and quote before you decide. Sending an enquiry carries no obligation. The introductory 20-minute call is free.
The short version
- Ask about process before tools, proof before promises, and handover before lock-in.
- Assess workflow fit, clear ownership and measurable outcomes.
- Good answers have numbers, limits and an exit plan. Bad answers have adjectives.
- Walk away from guaranteed ROI, tool-pushing, jargon walls and hidden pricing.
- The best people make themselves unnecessary. Hire for the handover.
FAQ: hiring an AI consultant or coach
How much should an AI consultant cost?
Tell me what you want to achieve. I’ll recommend a suitable approach and provide a clear scope and quote before you decide. Sending an enquiry carries no obligation. The introductory 20-minute call is free.
Do I need a consultant or a coach?
Hire a consultant when you want analysis or a system built for you. Hire a coach when you want to learn to run AI yourself, so the capability stays after they leave. For most small businesses and individuals, coaching pays off faster — you're building a skill, not renting one. Some people do both: coaching to get capable, then a builder for the automations you don't want to own.
What questions should I ask an AI consultant before hiring them?
Start with three: walk me through a real result with numbers, what won't AI fix in my business, and what happens when you leave. Those answers tell you whether you're getting proof or promises, honesty or hype, and handover or lock-in. Then cover who does the work, data handling, total price, the week-one plan and how success will be measured.
What are the red flags when hiring an AI consultant?
Guaranteed-ROI promises, one tool pushed for every problem (especially if they resell it), jargon they can't translate into plain English, no curiosity about how your work actually happens, and a price that only appears after a paid discovery engagement. One of these is a reason to slow down; two or more is a reason to walk away.