Professional Services Firms Are Using AI Everywhere and Measuring It Almost Nowhere
AI use in professional services doubled in a year. Only 18% of firms track ROI. That gap is about to become a very expensive problem.
AI adoption inside professional services firms nearly doubled in a single year. Organization-wide deployment went from 22% in 2025 to 40% in 2026, across more than 1,500 respondents spanning 27 countries. That's a real and fast shift. The problem sitting directly underneath it: only 18% of those same organizations actually track what any of it is returning.
Read that twice. The profession bought AI at scale and has almost no framework for knowing whether it was worth it. That's not a technology failure. It's a management failure, and it's one the industry is about to be forced to confront.
The ROI Gap Is Now a Boardroom Problem
The numbers here are blunt. A 2026 CEO survey found 56% of chief executives report neither increased revenue nor decreased costs from AI over the prior 12 months. Only 12% report achieving both. Meanwhile, 40% of professionals say they don't even know whether their firm measures AI return on investment at all.
So you have a situation where individual usage is racing ahead, 74% of professionals now use AI several times a week, and 44% use it multiple times a day, while institutional accountability sits nearly empty. Firms are spending real money on tools, training, and integrations, and most of them are flying blind on outcomes.
The metric of 2025 was "users." The metric of 2026 is "auditable outcomes." Firms that haven't made that shift yet are already behind.
Why Professional Services Gets Hit Hardest
Every industry faces the AI ROI measurement problem. Professional services faces it worst, because the business model itself is the casualty.
Time-based billing built an entire economic structure around hours. AI compresses hours. When a lawyer using Harvey AI or Westlaw Precision can complete a research task in 40 minutes that used to take four hours, the billable hour model doesn't just bend. It breaks. The productivity gain and the revenue hit land at the same time.
This is most visible in legal. AI-powered eDiscovery tools that large firms once held as a size-based advantage are now accessible to mid-size practices. Document review at scale is no longer a differentiator. The work gets done faster, but the question of who captures that value, client, firm, or nobody, remains genuinely unresolved across most of the industry. For a practical look at how smaller practices are handling this, how solo and small-firm lawyers are actually using AI in 2026 covers the ground-level reality well.
The same dynamic plays out in accounting, consulting, and any knowledge-intensive field where output used to be priced by time spent rather than value delivered.
What the Data Actually Shows About Usage
The usage picture is more nuanced than the headline adoption numbers suggest.
Individual usage runs well ahead of organizational deployment. While 40% of firms have organization-wide AI in place, 74% of professionals report using AI tools several times weekly on their own. That gap, between what firms have formally deployed and what people are actually running day-to-day, is a governance problem waiting to surface. Tools used without formal oversight don't get measured, don't get audited, and don't get caught when they go wrong.
On agentic AI specifically: only 15% of organizations currently use it, but 53% are either planning or actively considering it. By 2030, 77% of professionals expect agentic AI to be central to their workflow. For context on what agentic AI actually looks like when it runs into real-world friction, the Gartner report on $234 billion in enterprise SaaS spending at risk is worth reading alongside this.
More than 80% of current users engage with AI tools weekly. Over 90% expect AI to become central to their workflow within five years. The direction is not in dispute. The measurement and pricing infrastructure is.
The Client Disconnect Makes It Worse
Two-thirds of corporate clients say they want their outside professional services firms to use AI. Fewer than 20% mandate it. That gap creates a specific kind of confusion: firms get mixed signals, professionals receive conflicting guidance, and nobody sets clear expectations about AI disclosure, pricing adjustments, or quality standards.
Most professionals believe their firms should initiate clearer conversations with clients about AI use. Most firms haven't done it yet. That's a trust problem building in slow motion, and it's going to accelerate as clients start asking pointed questions about where their fees are going if AI is doing the work.
The Measurement Frameworks That Actually Work
The firms in the 12% that do report measurable returns share a common trait: they moved past usage surveys and into what some analysts are calling "time intelligence", tracking actual task completion times, comparing pre- and post-AI workflows, and pricing outputs rather than inputs.
That means:
- Baseline before you deploy. If you don't know how long a task takes without AI, you can't measure the delta with it.
- Define auditable outcomes. "Better work" isn't a metric. "Contract review time down 60%, error rate down to X%" is.
- Separate individual tools from firm-wide impact. An associate using a personal AI tool that the firm doesn't track is invisible to any ROI calculation.
- Revisit pricing models explicitly. Firms hoping this resolves itself are going to find clients doing it for them.
For firms in regulated industries, there's an additional layer: governance documentation. By 2026, clients in financial services and healthcare are increasingly expecting auditable AI delivery policies, not just assurances that AI is being used responsibly. The Bank of England's move to put agentic AI at the top of its supervisory agenda signals where regulatory pressure is heading, and professional services firms serving those clients need to be ahead of it.
What Firms Should Do Right Now
The window for treating AI ROI measurement as someone else's problem is closing. Here's what firms that want to stay ahead of this should actually be doing:
1. Audit current tool use across the firm. You probably have more AI running inside your organization than your IT department knows about. Individual subscriptions, browser plugins, and embedded features in existing software all count.
2. Build a tracking layer before the next tool purchase. Require ROI measurement criteria as a condition of any new AI investment. If you can't define what success looks like before you buy, don't buy.
3. Have the pricing conversation with clients proactively. Clients who discover their firm has been billing full hourly rates for AI-assisted work without disclosure are going to be angry. Clients who were consulted on it are partners in the transition.
4. Document governance, not just usage. Which tools are approved? Who reviews AI-generated outputs before they go to clients? What happens when something goes wrong? These questions need answers before regulators or clients ask them.
5. Consider value-based pricing pilots. Some firms are already testing project-based or outcome-based fee structures for AI-heavy work. The firms doing this now will have real data and client relationships built around the new model before the rest of the industry is forced into it.
Tools like Luminance for legal document analysis or Ironclad for contract management increasingly include usage reporting that can feed into internal ROI tracking. That's worth factoring into tool selection if measurement is a priority, which it should be.
The broader AI industry is wrestling with the same accountability gap, and the stakes keep rising. As we covered in the AfterQuery valuation story, the money flowing into AI infrastructure assumes that enterprise adoption translates to enterprise value. Professional services firms are the canary in that particular mine. If they can't close the measurement gap, the whole thesis gets harder to defend.
The honeymoon phase is over. The firms that figure out AI accounting now are going to be in a structurally better position than the ones still running on optimism by the time 2027 reporting cycles hit.
