Every owner I talk to is using AI now. That is no longer the question. The question is why so few of them can point to a dollar it made.
A new study puts hard numbers on that gap. SPI Research's 2026 Impact of AI on Professional Services report, released October 20 and based on 245 professional services organizations, found AI running on 43.6% of projects, nearly double the 23.1% from a year earlier. Adoption is not the problem anymore. Conversion is.
The short answers up top
What is the adoption gap? It is the distance between using AI and profiting from it. In the SPI study, 84.8% of firms report real value from AI through time savings. Only 29.9% report value through revenue. Same tools, same year, wildly different outcomes.
Why doesn't saved time become revenue? Because a saved hour is not a result. If the workflow around it never changes, the hour dissolves back into the same week. Revenue shows up when the freed capacity is aimed somewhere on purpose, and when somebody is counting.
What closes the gap? Depth and measurement. The firms on the right side of these numbers run AI across most of the operation, on top of systems that are wired together, and they track outcomes in dollars. SPI calls the pattern a Readiness Premium. We would put it more plainly: deployed beats advanced.
The numbers, without the spin
Give SPI credit for asking the uncomfortable version of the question. Not "do you use AI," but "where did the value land."
Time savings: 84.8% of firms. Revenue generation: 29.9%. And the finding that explains the other two: 36.8% of firms do not formally track any AI outcomes at all. Only 19.1% track what AI does to billable utilization, the number closest to money in a services business.
SPI's R. David Hofferberth said it in one line: "Saving time isn't the same as increasing revenue." The opportunity, in his words, is pairing AI with reliable data, trained people, and measurable outcomes so the productivity turns into business value.
That pairing is the whole game. Most firms bought the first half of the sentence.
Where the saved hours actually go
Talk to owners who feel stuck and a pattern repeats. Nobody lied to them. The tool works. The hours really do get saved. Then three things happen.
First, the hour evaporates. An hour saved inside an unchanged workflow gets refilled with more of the same work, because nothing redesigned what the freed time is for. Capacity becomes revenue only when it is aimed: more clients served with the same team, turnaround times you can quote tighter than the firm across town, follow-up that used to slip on busy weeks now happening every single time. If nobody makes that call, the calendar just fills back up.
Second, the tool sits beside the work instead of inside it. A chat window here, a summary generator there, maybe an agent answering the website. Meanwhile the systems the business actually runs on, the CRM, the job board, the books, never feel any of it. AI that lives next to the operation can save minutes. AI wired into the operation changes what the operation produces. This is also why the front door matters now: agents are already visiting your website and reading your business whether your systems are ready for them or not.
Third, nobody is counting. More than a third of firms track nothing, which means the AI program runs on anecdotes. It feels modern. It feels busy. The revenue line has no opinion about feelings, and without a number attached to the work, the first budget squeeze kills the program and nobody can say what was lost.
The firms on the other side of the gap
The same study shows what the other side looks like, and it is not "firms with better models."
Firms running AI on more than 80% of projects deliver on time 90.7% of the time. Firms under 20% sit at 76.3%. Firms running complex agents or AI-first operating models report services EBITDA margins of 20.5%, against 13.3% for firms at the chat-assisted stage.
Read those three numbers together and the story is consistent. The payoff does not come from the cleverness of the tool. It comes from how much of the operation the tool is allowed to touch, and whether the foundation underneath it, connected systems, clean data, people who are trained on it, can hold the weight. A chatbot bolted onto a messy operation produces a messy operation with a chatbot. A system built into the way the business already runs compounds.
There is a customer angle here too. In the same study, 59.9% of firms say their clients now expect AI-enabled delivery, up from 45.0% a year ago. Standing still is becoming something your customers can see.
What closing the gap looks like at your size
Your business may not look like a 245-firm survey of professional services organizations. The pattern scales down fine. Closing the gap is four decisions, none of them about which model is newest this month.
Wire the work together first. AI earns its keep inside the flow of the business, quoting, scheduling, delivery, billing, follow-up, not in a separate tab somebody opens when they remember. That usually means connecting the systems you already pay for before buying anything new.
Give the business one memory. Every firm runs on knowledge that lives in the owner's head, in text threads, in somebody's notebook. We build that layer as a Company Brain: the company's own policies, past jobs, pricing logic, and hard-won know-how in one place every system can draw on, so the AI answers from your business instead of from the internet, and your team stops interrupting the owner for answers the company already has.
Decide the dollar outcome before you build. Pick the number the system has to move, jobs quoted per week, days to invoice, repeat bookings, and write it down. Then track it weekly, the way you track cash. The 36.8% who track nothing are not careless people. They just never made measurement part of the build. Make it part of yours.
Keep a person responsible after launch. Systems drift, tools update, teams turn over. The firms in the 29.9% treat AI like operations, with an owner, a review rhythm, and a human checking the work. The rest treat it like a purchase, and purchases depreciate.
None of this requires the most advanced model on the market. It requires the decision to finish the job: architecture first, measurement always, and a system that runs the whole operation instead of decorating the edges of it. If you are pricing that kind of build, our breakdown of what AI automation actually costs a small business uses real numbers instead of vibes.
Book a free 30-minute diagnostic at krastor.com/diagnostic and leave with a map of where your AI hours are going and what to wire up first, whether we build it or not.
Frequently asked questions
What is the AI adoption gap?
It is the distance between using AI and profiting from it. SPI Research's 2026 study of 245 professional services firms found AI running on 43.6% of projects, with 84.8% of firms reporting value through time savings but only 29.9% reporting value through revenue. Adoption nearly doubled in a year. Revenue barely moved.
Why does AI save time without increasing revenue?
Because a saved hour is not a result. Inside an unchanged workflow, the freed hour gets refilled with the same kind of work, the tool sits beside the operation instead of inside it, and nobody tracks the outcome in dollars. Time savings only turn into revenue when the work around them is redesigned on purpose: more capacity aimed at clients, faster turnaround, follow-up that used to slip now happening every time.
How many firms track whether AI is working?
Fewer than you would expect. In the SPI Research study, 36.8% of firms do not formally track any AI outcomes, and only 19.1% track AI's impact on billable utilization. What is not measured in dollars gets managed by anecdote, which is how a firm can feel busier and more modern while its revenue line stays flat.
What separates firms that profit from AI from firms that just use it?
Depth and foundations. Firms running AI on more than 80% of projects report 90.7% on-time delivery, against 76.3% for firms under 20%. Firms running complex agents or AI-first operating models report services EBITDA margins of 20.5%, against 13.3% at the chat-assisted stage. The winners wired AI into the operation and measure what it returns. Deployed beats advanced.
Continue reading
- AI Agents Are Already Visiting Your Website. You'd Never Know.
- Grok's Bot Just Got Its Own Email Address. Here's What It Means for Your Business.
- What AI Automation Actually Costs a Small Business (Real Numbers, No Pitch)
- Book your free 30-minute diagnostic
Common questions
- What is the AI adoption gap?
- It is the distance between using AI and profiting from it. SPI Research's 2026 study of 245 professional services firms found AI running on 43.6% of projects, with 84.8% of firms reporting value through time savings but only 29.9% reporting value through revenue. Adoption nearly doubled in a year. Revenue barely moved.
- Why does AI save time without increasing revenue?
- Because a saved hour is not a result. Inside an unchanged workflow, the freed hour gets refilled with the same kind of work, the tool sits beside the operation instead of inside it, and nobody tracks the outcome in dollars. Time savings only turn into revenue when the work around them is redesigned on purpose: more capacity aimed at clients, faster turnaround, follow-up that used to slip now happening every time.
- How many firms track whether AI is working?
- Fewer than you would expect. In the SPI Research study, 36.8% of firms do not formally track any AI outcomes, and only 19.1% track AI's impact on billable utilization. What is not measured in dollars gets managed by anecdote, which is how a firm can feel busier and more modern while its revenue line stays flat.
- What separates firms that profit from AI from firms that just use it?
- Depth and foundations. Firms running AI on more than 80% of projects report 90.7% on-time delivery, against 76.3% for firms under 20%. Firms running complex agents or AI-first operating models report services EBITDA margins of 20.5%, against 13.3% at the chat-assisted stage. The winners wired AI into the operation and measure what it returns. Deployed beats advanced.