Generative AI adoption inside professional-services firms has nearly doubled in the last year. 40% of professionals now say their organization actively uses it, up from 22% just twelve months ago, a real, fast shift in how legal, accounting, and consulting work actually gets done. Source

That number alone would normally be the whole story. It isn’t. Ask the same group of professionals a much simpler follow-up question, is anyone actually tracking whether any of this is paying off, and the story changes completely.

The Measurement Gap Nobody Budgeted For

Only 18% of professionals say their organization tracks return on investment on its AI tools, and that’s mostly through informal, internal metrics rather than anything rigorous. Another 40% say they genuinely don’t know whether ROI is being measured at all. Source

Sit with that for a second. Adoption nearly doubled. Measurement didn’t move with it. Two out of every five firms using AI right now have no idea if it’s actually working, they’ve simply adopted it and moved on, the same instinct that drove a decade of software purchases nobody ever went back and audited.

Where the Money Actually Shows Up First

Not every corner of professional services is flying blind. Audit and accounting practices are seeing the fastest payback of any professional-services use case, 180 to 250% in first-year returns, largely because their work is structured, repeatable, and increasingly priced on a fixed basis rather than by the hour. Source

That structure matters more than it sounds like it should. When a task has a clear before-and-after, categorizing a transaction, reconciling an account, drafting a routine filing, it’s genuinely easy to measure whether AI made it faster or better. Bookkeeping-specific automation is already delivering a 67% reduction in transaction categorization time industry-wide, with some practices reporting up to 85% time savings on the most routine data entry work. Source Firms in that lane didn’t get lucky. They picked use cases where “did this actually work” has a real, checkable answer.

Legal tells a slightly different story. Adoption is real and heavily concentrated in a handful of tasks, legal research (80%), document review (74%), document summarization (73%), and drafting briefs or memoranda (59%). Source Tax and accounting show a similar pattern, tax research (69%), summarization (57%), review (55%). These are exactly the tasks where output can be checked against a known standard. That’s not an accident, it’s the actual blueprint every firm should be borrowing from.

Adoption Without Strategy Is a Coin Flip

Here’s the number that should actually change how a firm approaches its next AI decision. Organizations with a formal AI strategy are roughly three times more likely to achieve a real, positive return than organizations without one. Source Not marginally more likely. Three times.

That statistic quietly reframes the entire “should we adopt AI” conversation most firms have already had and moved past. The real conversation, the one almost nobody’s having, is “do we actually have a way to know if this is working,” and right now the honest answer for most firms is no.

This Isn’t Just a Professional-Services Problem

The same pattern, adoption running ahead of proof, isn’t unique to law firms and accounting practices. It shows up in a different costume across every vertical iSC works in.

In Home + Trade Services, the equivalent gap isn’t measurement, it’s speed. Weekend unanswered-call rates jump to 41%, more than double the weekday rate, and most trade businesses have no system tracking how many of those calls they’re actually losing in the first place. Source

In Beauty + Wellness, medspas using AI booking tools posted 5% sales growth versus 1% for non-users, yet most of that lift comes from data owners already had sitting in software they already pay for, unused for exactly this purpose. Source

In Restaurant + Caterer, 86% of operators are already comfortable using AI, but 89.7% of caterers still rely on word-of-mouth as their primary growth channel, comfort with the tool hasn’t yet translated into a measured strategy for using it. Source

And in Corporate + B2B, 64% of businesses using AI chatbots report more qualified leads coming through, but only 55% of the sales and marketing leaders running those chatbots say lead quality actually improved, meaning nearly half of current deployments aren’t tuned well enough to prove their own case. Source

Different industries, same root issue. Somebody adopted the tool. Almost nobody built the scoreboard.

What Proof of Value Actually Looks Like

None of this requires a data science team or a six-month audit. The firms already seeing real returns share three habits, not budgets.

First, they picked a starting use case with a clean before-and-after, something with a genuinely measurable output, not a vague productivity claim nobody can pin down. Second, they wrote down what “working” would actually look like before they turned the tool on, not after the fact when it’s convenient to grade on a curve. Third, they checked back on a real, recurring schedule, not once at the one-year renewal when it’s already too late to course-correct cheaply.

That’s genuinely it. It’s a habit, not a system purchase, and it’s the difference between the firms quietly compounding a real advantage right now and the ones that will spend next year’s budget cycle trying to explain what last year’s AI spend actually did for them.

The Question Worth Asking This Week

If your firm adopted an AI tool sometime in the last year, ask the person who championed it one question: what changed because of it, specifically, and how do we actually know? If the answer is a shrug, a vague “it’s helped,” or a sentence that starts with “I think,” that’s not a failure of the tool. It’s a gap in how the rollout was measured from day one, and it’s genuinely fixable without ripping anything out.

Adoption was never the hard part, every vertical iSC serves has already proven that part is easy. Proof is the actual differentiator now, and building it in from the start is a lighter lift than most firms assume, once someone actually sits down to define what success looks like before the next renewal instead of after it.

Not sure whether your current AI tools are actually earning their keep? Book a quick call