Before you hire an AI advisor, ask yourself these eight questions

Everyone's an AI expert this year. Most of them aren't.

I keep seeing the same pattern. A marketing director who's spent three months with ChatGPT runs a webinar on "how to write the perfect prompt." A digital agency bolts "AI transformation" onto its services menu because it fits its existing wheelhouse. A consultant with no governance background tells you to paste your client call recordings into a free AI tool to extract insights, mentions GDPR once at the end, and moves on.

None of these people are necessarily bad at what they actually do. But there's a difference between knowing how to use AI and knowing enough to guide someone else through it without creating risk. Right now, that difference isn't showing up anywhere obvious.

The real risk isn't sounding robotic or missing productivity gains. It's that one piece of wrong advice, delivered with total confidence, can expose your client data, create compliance gaps, or leave you managing a half-built implementation that nobody actually understands.

Here are some tips to help you assess whether the person asking to guide your AI work has earned the right to do so.

What to ask before you hire them

1. Can they show you work they've actually done, not just content they've published?

There's a difference between someone who's used AI in their own business and someone who's implemented it in somebody else's. Ask for a specific client situation: what did they build, what changed, and what broke or needed revision?

If they arrive with a workshop deck they use everywhere, that's not diagnosis. That's a template rotating through every client.

2. Have they worked with a business like yours, or with comparable risk?

Someone who's optimised social media workflows for a coaching solopreneur is not automatically qualified to guide a healthcare clinic or a legal firm.

That doesn't mean they need a long client list in your exact sector. But they should be able to show that they've worked in an environment with comparable data, regulatory, operational or reputational risk, and explain what changed because of it.

3. Are they teaching judgement or selling a static set of prompts?

If their offer centres on "here's the perfect prompt for X" or a pack of pre-built templates, I'd ask what happens when those prompts stop working.

Good prompting still matters, but it isn't a collection of magic commands. Effective AI use now depends much more on context, iteration, examples, constraints, feedback and knowing how to judge the output.

A shallow educator teaches you which words to type. A credible one teaches you how to think about what the tool is doing, where it can fail and what to do next.

4. Can they competently cover data handling, confidentiality, IP, bias and vendor terms, or do they treat those as someone else's problem?

This is where bad advice compounds fastest: if someone tells you to paste client data into a free tool and skips what happens to it, what the terms actually say, or when you need legal input, they're showing you the edge of their knowledge.

A competent guide understands the risks well enough to spot them, and knows when to bring in a specialist rather than pretending to be one.

5. When was their material last substantively updated, and what changed?

AI advice dates quickly. Models change. Products change. Terms change. Regulation changes. Capabilities that were cutting edge a year ago can become standard, disappear entirely or work differently.

Ask when their course, workshop or methodology was last materially updated, what they changed and why.

If the answer is "it's still relevant", that's worth probing. Relevant to what, exactly?

6. Do they start with your work, or do they arrive with a generic workshop?

A competent guide diagnoses before they prescribe: they ask about your constraints, your actual use cases, your existing tools, what your team can already do and what happens after the initial implementation.

Who owns the workflow? Who maintains it? What happens when the model changes? Is it documented well enough for somebody other than the consultant to understand?

If they hand you a one-size-fits-all playbook on day one, they're not diagnosing. They're applying a template.

7. Are they paid to recommend a particular tool, platform or stack?

If they're an affiliate for a specific automation tool, or they work for a platform they're recommending, that's not automatically disqualifying. But it matters.

They should disclose it, and you should be able to tell whether the recommendation is genuinely the best fit for your business or simply the thing they know how to sell.

8. Can they distinguish what they know from when you need someone else?

Someone with real depth knows where they stop. They can say, "I can help you design the workflow", and separately, "you need a data protection specialist here", or, "that's a legal question."

If someone presents themselves as the answer to every part of the problem, they're not competent. They're confident beyond their scope.

Why this matters

One piece of bad advice, delivered with authority, can put your client data at risk. You end up with recordings sitting in someone else's cloud. You agree to terms you didn't read. You inherit an implementation nobody on your team understands.

The people who can actually advise you on AI are the ones who've lived through the messy parts: the failed pilots, the moment the governance rules conflicted with the tool's terms, the discovery that the "quick win" created more work than it solved. That's where judgement comes from, not from having the loudest opinion about AI, a prompt pack, or a completed course with "AI" added to a LinkedIn headline.

Before you hire someone to guide your AI work, look for evidence over confidence.