How to Tell If Someone Actually Knows What They're Doing in AI

There's a lot of confidence in the AI market right now. There's not nearly as much competence.

For a business owner, this creates a real risk. You can spend meaningful time and money on someone who sounds credible — but can't deliver anything durable.

So the question becomes: how do you tell the difference?

01 · Curiosity, not credentialsThe best signal isn't a certificate.

It's behavior. People who are actually good at AI:

Constantly test new tools
Break things on purpose to understand limits
Change their opinions as the technology evolves

If someone presents AI as a fixed playbook, they're already behind.

02 · Credentials in contextNecessary but never sufficient.

There are plenty of courses, certifications, and "AI experts" right now. None are meaningless. None are sufficient. What you actually want to understand:

Have they worked inside real businesses?
Have they dealt with messy data and unclear processes?
Have they shipped something that people actually use?

AI in theory is easy. AI in operations is not.

03 · Evidence of applied workThe clearest filter.

Ask for:

Specific workflows they've improved
Before-and-after metrics — time saved, errors reduced, cost impact
What broke during implementation, and how they fixed it

If everything sounds clean and linear, it probably wasn't real. Real work is messy. Good operators are comfortable talking about that.

04 · Tool-centric thinkingThe most common red flag.

You'll hear it as:

"We'll use this model."
"We'll connect these platforms."
"We'll build a custom GPT."

Tools change constantly. Workflows persist. The right person talks about process design, decision points, and where humans stay involved versus where they don't.

If the conversation is mostly about tools, you're not talking to a strategist. You're talking to a technician — or worse, a salesperson.

05 · The trust testLook for honest framing.

Ultimately, this comes down to trust — but not in a vague sense. You're looking for someone who admits uncertainty, frames tradeoffs clearly, and doesn't oversell outcomes.

A clear thinker says…

"This part is well-suited to AI; this part isn't."
"Here's what could go wrong, and how we'd catch it."
"We'd need a human in this loop."
"This will probably take two iterations to get right."
"I don't know yet — I'd want to test it."

Walk away when you hear…

"AI can do anything you want it to."
"This will pay for itself in 90 days, guaranteed."
"You won't need that team anymore."
"Just let the model handle it end-to-end."
"It just works."

AI has real upside — but also real limitations: it can hallucinate, it can degrade without oversight, it can introduce risk if misapplied. If someone doesn't talk about these things, they're either inexperienced or not being honest.

If you walk away thinking "this feels clearer and more grounded," that's a good sign. If you walk away thinking "this sounds like magic," it's not.