What the AI consulting boom gets wrong

Key takeaways

• Demand for AI help is exploding, and most of the money will be spent on advice rather than working systems.
• The buyers driving the demand aren't shopping for a model. They're shopping for someone to close the gap between what AI can do and how their operation actually runs.
• The gap is closed with vertical specifics: the property software, the intake workflow, the compliance question. General AI expertise doesn't survive contact with any of it.
• Four questions to ask any firm before hiring one. Including us.

Everyone is an AI consultant now

You don't need a market report to see the boom. Every marketing agency has rebranded around AI. Every software vendor has bolted "AI-powered" onto the product page. LinkedIn has filled up with prompt consultants, automation gurus, and transformation practices spun up in a weekend. Under all the noise there's a real signal: an enormous number of businesses have decided, at roughly the same time, that they need help with this.

They're right. What most of them will buy, though, is advice. Slide decks about AI readiness, maturity models, strategy workshops. The deliverables land, everyone nods, and six months later the operation runs exactly the way it did before. The boom is real. Most of it will produce nothing.

What the demand actually is

Listen to what buyers are actually asking for and a pattern shows up fast. Nobody calls us asking for a large language model. The calls sound like this instead: our phones go unanswered after six. Leads sit in an inbox all weekend. Every lead gets retyped into the CRM by hand. Month-end reporting eats a full week. Our maintenance history lives in one person's memory.

None of those are AI problems. They're operations problems that AI happens to be very good at now, and that distinction decides who can actually help. The person who needs to close that gap doesn't need a smarter model. The models are already good enough for almost everything on that list. They need someone who can land the model inside the operation: wire it to the property software, put a human approval in the right place, keep a paper trail, and leave the team owning the result.

Why generalists miss

Here's what the generalist pitch never has to deal with. The property manager's leads live in Buildium or AppFolio. The brokerage runs on the MLS and a CRM the brokerage picked years ago and can't leave. The trades business books jobs through a phone line that rings to a cell after hours. The clinic can't send patient data through a hosted automation platform in another jurisdiction, and someone eventually has to answer, in writing, where the data goes.

Every one of those details is invisible from the strategy deck and decisive in production. The work isn't "apply AI to the business." The work is a few hundred lines of careful integration code, a prompt tuned against real messages, an approval step where the operator wants it, and documentation the team will actually read. Firms that have shipped inside a vertical carry that knowledge with them. Firms that haven't will spend your budget acquiring it.

That's the quiet reason the boom will disappoint most of its buyers. The bottleneck isn't AI expertise. It's operational specificity, and operational specificity doesn't scale the way a consulting practice wants it to.

How to buy well in a gold rush

We're not neutral here. We're one of the firms in the boom, and this essay is on our own website. So judge every firm, including this one, by the same four questions:

1. Have they shipped in your vertical? Not "worked with businesses like yours." Shipped: working systems, running today, inside the software you actually use. Ask what broke and what they did about it. Firms that have shipped have answers with specifics in them.

2. Is the scope fixed and the demo weekly? Open-ended discovery is a billing model. You want an agreed scope, an end date, and a working demo every week against your own data, so a wrong turn costs you a week instead of a quarter. (This is [the engagement structure we use](/blog/audit-build-operate-model), written up in full.)

3. Do you own everything at handoff? Source code, prompts, credentials, documentation. If any of it stays with the firm, you're not buying a system, you're renting a dependency.

4. Who, by name, runs it after they leave? If the firm doesn't insist on naming an internal owner before the build starts, they're planning a demo, not a deployment. It's the difference we wrote about in [why most AI pilots die after the demo](/blog/why-most-ai-pilots-die-after-the-demo).

The point

The boom will sort firms into two piles: the ones selling proximity to AI, and the ones accountable for systems that run without them in the room. The first pile is bigger and always will be. The second pile is where the money you spend turns into hours you get back.

Ask the four questions. Whoever you ask them of.