How to vet an AI consultant before you hire one

Cody Reppert · Field notes

AI consulting is new enough that there is no license, no board, and no standard to hold anyone to. That means the burden is on you to ask the right questions before you sign anything. Here is what separates a real operator from someone riding the wave.

They can't say what a person is responsible for

Ask where AI stops and a person takes over. If the answer is vague, or the honest answer is "nothing catches it," that is the whole conversation right there. A real system has a person reviewing anything that carries risk before it reaches a client. If nobody can tell you where that person sits in the process, there isn't one.

There's no plan for after launch

Models change, your business changes, and a system built once and never touched again degrades quietly. If the engagement ends the day something gets deployed, with no support window and no plan for upkeep, you have bought a demo, not a system.

The pitch is the demo, not your workflow

A slick demo proves a tool can do something in general. It does not prove it fits how your team actually works. Watch for whether someone is asking real questions about your process before they start building, or just showing you what the technology can do in the abstract.

Pricing is hourly with no scope

Hourly billing on an open-ended AI project puts the risk on you. A fixed price for a defined outcome puts the risk on the person building it, which is where it belongs. If nobody will commit to a scope and a price, that tells you something about how confident they are in delivering it.

They can't say what they won't do

Anyone doing this work responsibly should be able to name their own limits without hesitating: what they won't automate, where they stay out of regulated decisions, what stays human no matter what. If the pitch is "AI can do anything," that confidence should make you more cautious, not less.