84.8% Say AI Saves Time. 29.9% Say It Pays.

SPI Research surveyed 245 professional services firms between July and September of this year, and two numbers in the results belong right next to each other. 84.8% of the firms said AI saves them time. Only 29.9% said it generates revenue, according to the summary published by the report's sponsor.

SPI calls the distance between those two numbers "the leak." It matters more now than it did a year ago, because AI use in this work nearly doubled in that time: AI now touches 43.6% of projects at these firms, up from 23.1% in 2025.

SPI's sample is professional services organizations broadly, from IT and management consultancies to the services teams inside software companies. It isn't a survey of boutique advisors or senior independents. But if you run a small advisory practice and you've started using AI in your own work, the leak is the most important part of the report for you, and it has a simple cause.

The savings have to go somewhere

If you bill by the hour and AI takes hours out of the work, your invoice gets smaller. The research that took you twelve hours now takes four, and you bill four. The client gets the same deliverable for less, which is good for the client and a pay cut for you, even though you did nothing wrong except get better at your job. Spread that across a year of engagements, and the tool that made you faster has quietly lowered your income, while your clients get a better deal they never asked you for.

Erik Brynjolfsson and Georgios Petropoulos at the Stanford Digital Economy Lab lay out why this happens in a pricing framework they published in August. When consultants use AI inside their own work, they argue, the client mostly can't see it. The deliverables look the same. What changes is that "hours are no longer reliable proxies for effort," so time-based pricing becomes a worse instrument than it was, and clients on time-based contracts "pay for hours that no longer correspond to the work being done."

That's the leak from the other side of the table. A firm that bills hours and then removes hours from its delivery has handed its efficiency gain to the client, one invoice at a time.

Most independents have already left the pure hourly model

None of this means every advisor is billing by the hour. A preliminary 2026 survey from Reinvention Academy found that only 30% of consultants and coaches primarily charge by the hour, down from 36% two years ago, and 55% now charge by the project.

So the honest version of the problem is narrower and more interesting. Plenty of project fees are still an hours estimate wearing a nicer outfit. If your price for a competitive analysis is built from how many hours you think it'll take times your rate, AI shrinks the estimate the same way it shrinks an hourly invoice, just one proposal later. The leak doesn't care what the line item is called. It cares what the number is made of.

Cheaper work can mean more work, if the price lets it

There's a real counterargument, and it's worth taking seriously. Mike Piehl, a services leader interviewed by theCUBE Research this month, described what he called a labor-demand paradox: "The labor demand on an individual project is coming down. However, because the overall price points are coming down as well and the value delivery is going up, it actually means that we have more client demand as a result."

That's one leader's experience, not a survey finding, but the logic holds. If AI makes a project cheaper to deliver, projects that used to be too small to bother with become worth doing, and total demand can rise. The catch is that the paradox only pays off if your pricing lets the efficiency turn into more work at a healthy margin. If every saved hour comes straight off the invoice, you're doing more projects to stand still.

SPI's data points in the same direction. Firms running complex agents or an AI-first model reported EBITDA of 20.5%, against 13.3% for firms using AI only as a chat tool or not formally at all, per the sponsor's summary. That's an association, not proof that AI caused the difference, and it says nothing about boutiques specifically. It does suggest the firms getting paid for their AI work are doing something different from the ones only saving time with it.

What to price instead

Clients are starting to notice on their own. The same Stanford paper says sophisticated clients are already pushing back on time-based contracts with firms that use AI internally, which means the leak isn't only a margin problem waiting to happen. It's a conversation some of your clients are going to start whether you're ready or not.

The paper itself is useful here because it doesn't just say "stop billing hours." It maps five pricing models – time-based, project-based, access-based, outcome-based, and hybrid – onto two questions. How easily can the client see the outcome of your work? And how easily can they see what it cost you to produce it? Hourly billing works when the client can see your effort and can't easily judge the result. Outcome-based pricing works when the result is plain to everyone. Most advisory work sits somewhere uncomfortable in between, because the value of a good recommendation often isn't obvious until months later, and AI has just made your effort much harder to see.

Brynjolfsson and Petropoulos predict that internal AI use pushes consulting contracts away from time-based pricing and toward access-based or hybrid structures. Access-based pricing is the retainer: in their words, "a strategic advisor whose presence is the deliverable," where the unit of value is availability because no other unit is reliably measurable. They also note that time-and-materials holds up in highly specialized expert advisory, which is worth keeping in mind if your work is that kind.

For a senior advisor, I'd put it more simply. Your clients were never paying for your hours. They were paying for your judgment: what you notice in the material, what you decide matters, and your name on the recommendation when they take it to their board. Hours were a convenient way to measure that because they were roughly proportional to the work. AI broke the proportion. The judgment didn't get cheaper.

I price my own work this way. A Research engagement with me is a fixed price in the low four figures, scoped to one question, and the price doesn't move with how many hours the answer takes. The research tools I use keep getting faster, and if I billed by the hour, every one of those improvements would have come straight out of my income.

That's also what makes it safe to use help behind the scenes, whether that help is AI, a research partner, or both. You stay the face of the engagement, and the client is buying what only you can give them. The part that can be done faster should be done faster, and your price shouldn't go down when it is.

If you've been using AI for a few months and your revenue hasn't moved, check what your prices are made of before you check anything else. The day rate was always the wrong number to anchor on, and AI has made that a lot harder to ignore.