How to Choose an AI Consultant (Without Getting Burned)

There's no shortage of people calling themselves AI consultants right now. Some of them are brilliant. Some of them were doing something completely different eighteen months ago and pivoted when ChatGPT made the news. The problem is, from the outside, they look exactly the same.

I've been on both sides of this. I've been the consultant pitching solutions, and I've been the person inside a company evaluating consultants. The gap between a good engagement and a bad one is enormous. A good consultant saves you months and hundreds of thousands of dollars. A bad one burns both and leaves you with a codebase nobody can maintain.

Here's what I'd look for if I were hiring someone to do what I do.

Ask what they'll tell you not to build

This is the single biggest signal. A consultant who says yes to everything is either desperate for the work or doesn't understand the problem well enough to push back.

The best consultants I've worked with (and the standard I hold myself to) will tell you when AI isn't the right answer. Sometimes the problem is a data problem. Sometimes it's a process problem. Sometimes you just need a better spreadsheet. If the person across the table can't articulate when AI is the wrong tool, they're selling technology, not solving problems.

Look at what they've built, not what they've presented

Slide decks are easy. Production systems are hard. Ask to see something that's actually running. Not a demo, not a proof of concept, not a pitch deck with architecture diagrams. Something that handles real data, real users, real edge cases.

If they can only show you demos, that tells you something. Demos are the easy part. The hard part is what happens when the model hallucinates at 2 AM and there's no one watching.

Check if they understand your business, not just the technology

AI is a tool. It's a powerful tool, but it's still a tool. The consultant's job isn't to implement the most impressive technology. It's to solve your specific problem in the most effective way possible.

If someone spends the entire first conversation talking about models and frameworks and never asks about your customers, your revenue, your team, or your actual pain points, that's a red flag. The technology conversation should come after the business conversation, not before it.

Ask about failure

Every experienced consultant has stories about things that went wrong. Projects that missed the mark. Approaches that didn't work. If someone tells you they've never had a project fail or underperform, they either haven't done enough work or they're not being honest.

What matters isn't whether they've failed. It's what they learned and how they adjusted. The best consultants have strong opinions about what doesn't work because they've tried it.

Understand their process before you see their proposal

A good consultant has a repeatable process. Not a rigid one, but a structured approach to understanding your problem, evaluating options, designing a solution, and building it. If the process is "tell me what you want and I'll build it," that's a contractor, not a consultant.

You want someone who will push back on your assumptions, challenge your requirements, and tell you things you don't want to hear. That's uncomfortable. It's also how you avoid building the wrong thing.

Watch out for the "platform" pitch

Be cautious of consultants who arrive with a pre-built platform or framework that they want to deploy in your environment. Sometimes that's legitimate. Often it means they're going to force your problem into their solution rather than designing something that fits.

The best solutions are tailored. Not built from scratch every time (that would be wasteful), but designed around your specific constraints, data, and goals. If the proposal looks like it could have been written for any company in any industry, it probably was.

Pay attention to how they talk about cost

AI projects can be expensive. But "expensive" is relative. A $200K project that saves you $2M per year is cheap. A $50K project that produces nothing useful is the most expensive thing you'll ever buy.

Good consultants are transparent about cost from the beginning. They'll give you a realistic range, explain what drives the cost up or down, and help you make informed tradeoffs. If someone can't or won't talk about money until you've signed a contract, walk away.

If you're evaluating consultants right now and want a second opinion, I'm happy to have that conversation. No pitch, no obligation. Sometimes the most valuable thing I do is help someone figure out what questions to ask.