What an AI Consultant Actually Does (And What They Don't)

If you've talked to three AI consultants in the past year, you've probably gotten three completely different pitches. One came with a 60-slide deck about "AI maturity frameworks." One quoted you a fixed price to build a chatbot before they'd asked a single question about your business. One disappeared after the discovery call.

This is the state of AI consulting in 2026: a market full of practitioners who are either pure strategists, pure builders, or — most commonly — pure marketing. The actual job is something different.

What an AI consultant should actually do

A useful AI consultant does three things, in order: figures out what you should build, figures out whether it's worth building, and then either builds it or tells you who should.

The first part is harder than it sounds. Most businesses don't have an AI problem — they have a process problem, a data problem, or a hiring problem that AI happens to be one possible solution to. A consultant whose first instinct is "let's deploy a chatbot" is doing the equivalent of a contractor whose first instinct is "let's tear down the wall." Sometimes that's the answer. Most of the time it isn't.

The second part — figuring out whether it's worth building — is where most AI consulting engagements break down. Strategy decks rarely include the cost of operating a system 18 months in, the cost of training your team, or the cost of the alternative (which is often "do nothing and the business is fine"). A serious consultant runs that math before recommending the work.

The third part is the part most consultants skip. They hand off a roadmap, take their fee, and leave you to either hire developers or call another vendor. That handoff is where AI projects die.

The job isn't to recommend AI. The job is to figure out whether AI is the right answer to a real problem — and then either ship it or tell you who should.

What an AI consultant shouldn't do

The list is shorter, but it matters more.

A consultant shouldn't sell you a generic "AI transformation" that maps to your business roughly the way a horoscope maps to your day. Real engagements are specific: this customer, this workflow, this measurable outcome.

A consultant shouldn't charge senior rates for junior work. The most common pattern in AI consulting right now is a Big Five-style firm with a partner on the discovery call and a team of recent hires doing the actual work. You're paying $400/hour for someone learning AI on your dime.

A consultant shouldn't disappear after the strategy phase. If they can't or won't ship the implementation themselves, they should tell you who can — and not take a kickback for the referral.

A consultant shouldn't dress up an off-the-shelf tool as a custom solution. There are great off-the-shelf AI tools. There are great reasons to build custom. A consultant who can't articulate the trade-offs honestly is a consultant who's optimizing for their own margin.

What a real engagement looks like

Here's how we structure engagements at AIConsultants.co:

A free consultation, in person or virtual. We map what you're trying to accomplish, what you've tried, and where it stalled. If we don't think we're the right fit, we say so on the call.

A scoped engagement with a written statement of work. Fixed scope, fixed price, fixed timeline. Strategy work runs 2 to 6 weeks depending on depth. Implementation work runs longer and is scoped separately.

Senior practitioners on every engagement — the people on the consultation call are the people doing the work. We're a small firm by design.

A real handoff at the end. Documentation, training, credentials, runbooks. You can run the system without us. If you want us to keep operating it, that's a separate ongoing engagement; if you don't, you're not stranded.

The signs you're talking to a real one

Three quick filters when you're evaluating consultants.

Ask them what they've shipped recently. Specific projects, specific outcomes. If they can only describe frameworks and methodologies, they're a strategist who doesn't ship. That's fine — but you need to know that going in.

Ask them what they wouldn't take on. A consultant who'll take any project is a consultant who hasn't said "no" enough times yet to have a real specialty. The good ones have a clear sense of where they add value and where they don't.

Ask them what the engagement looks like 6 months in. Most AI projects fail in months 4-12 — when the novelty wears off, the original prompt engineer leaves, and the system needs ongoing care. A consultant who's never thought about month 12 hasn't been around long enough.

If you're working through this filter and want to talk to a team that's been through the cycle a few times, we'd be glad to have a real conversation about what you're trying to build.