Most leaders I talk to have already solved the question of whether their people will use AI. The tools are open on every desk. The harder question is whether all that usage is producing anything the business can point to. A report out this week says, for a lot of organizations, the answer is no. And it attaches a number to the cost.
On June 22, 2026, Thomson Reuters published its Future of Professionals Report 2026. The title says the thing directly: AI is ready, but firms are not. The technology works. The bottleneck is execution.
"Firms that are operationalizing AI are pulling ahead," said Steve Hasker, President and CEO of Thomson Reuters. "Those that aren't are starting to take on real risk."
That risk is not abstract. The report names two places it shows up. Clients leaving. Talent leaving.
The usage is high. The value is not.
Start with the split that defines the whole report.
of professionals now use AI tools every week, yet 91% of them believe their organization is falling short of what AI could deliver.
Source: Thomson Reuters, Future of Professionals Report 202674% of professionals now use AI tools every week. That is adoption most leaders would have called a win two years ago. People are in the tools. The habit is built.
But 91% of those same professionals believe their organization is falling short of what AI could deliver. Read that twice. Nearly everyone is using it, and nearly everyone thinks their company is leaving the real value on the table.
That is the disconnect the report circles around. Daily usage and actual return are two different things, and one does not produce the other on its own.
Here is where it gets practical for a leader. About a third of professionals say they use AI tools their organization never approved. Among people who think their organization is moving too slowly, that figure climbs to 41%. They are not waiting for the sanctioned path. They are building their own, quietly, with whatever tool is fastest.
And 41% say they lack access to professional-grade AI that meets their security and accuracy standards. So part of the unsanctioned-tool story is not rebellion. It is people reaching for something better than what the official stack gives them.
For anyone who owns risk at a professional services firm, that is the part to sit with. Work product touching legal advice, regulatory filings, and client deliverables is running through tools nobody vetted.
What the clients are saying
The report surveyed corporate clients, the people who hire law firms, accounting firms, and advisory shops. Their view is the sharpest part of the data.
78% of clients call AI-enabled quality improvements essential. They want the work to be better because of AI, not just faster.
Only 6% say their providers actually deliver that today.
Sit with the distance between 78 and 6. The demand is nearly universal. The delivery, by client perception, is rare. That space is where competitive advantage is being decided right now, and most firms have not stepped into it.
Then the consequence. 32% of corporate clients say they will reconsider which providers they use within the next 12 months, based on how those providers deliver AI. Of that group, a third have more than $1 million in annual work on the line.
Thomson Reuters models the total exposure at roughly $143 billion in U.S. legal and accounting revenue under active reconsideration. That figure is Thomson Reuters' own estimate built from the survey, not an independent audit. Treat it as their modeled number. Even discounted, it points the same direction: client relationships are being re-evaluated against AI delivery, and the money attached is large.
"Not all AI is created equal," Hasker said. "When outputs shape legal judgments, regulatory filings, or client advice, 'almost right' isn't good enough."
That line is the whole client problem in one sentence. The bar for professional work is not an impressive demo. It is work a professional can stand behind and put a name on.
What the talent is saying
The second cost is people.
24% of professionals would consider leaving their organization within two years over the distance between what AI can do and what their organization actually delivers. They are not frustrated by the technology. They are frustrated by watching their employer move slower than the tools allow.
62% say access to professional-grade AI would influence whether they accept a new role. AI capability has become a hiring factor, the way compensation and title have always been. The best people are now asking, in effect, will I be able to do my best work here, and is the tooling going to hold me back.
Put the two costs side by side. Clients reconsidering on one side. Your strongest performers screening their next job on AI access on the other. The organizations that close that distance keep both. The ones that do not are exposed on two fronts at once.
Why usage alone never closed the loop
So why does heavy usage produce so little value? The report points at execution without fully naming the mechanism. Here is the read worth carrying into your next leadership meeting.
Buying licenses and watching adoption climb feels like progress. It is the easy part to measure. But adoption is an input, not a result. A team can use AI every day inside a process that was never designed for it, and the output gets faster without getting better. Speed on top of a broken workflow just produces the same flawed work, sooner.
That is what the shadow-tool number is really telling you. When the official path is slow or weak, people route around it. They are not getting value from a designed system. They are improvising, one prompt at a time, with no shared standard for accuracy, security, or review.
The organizations pulling ahead did something different. They treated AI as a question about the work itself, not only the tools. They mapped the process first, decided where a human has to stay in the loop, set the quality bar, and then built the AI into that design. The order is the whole thing.
Design the workflow, then build the system to run it.
This is the practical core of moving up The 7 Levels of AI Proficiency, the measurable standard we use to describe how individuals and organizations grow from casual AI use toward real, dependable capability. The early levels are exactly what 74% of professionals already live in: using the tools, getting comfortable, building the habit. The higher levels are where this report says the value sits. People and teams designing the process, owning the judgment, and running AI as a managed part of how the work gets done. The jump from heavy usage to real return is the jump between those levels. It does not happen by buying more licenses.
What a leader can do with this
You do not need to solve all of this at once. A few questions point you in the right direction.
Find where your people are already routing around you
The shadow-tool number is a tell. If a third of your people are using unapproved AI, the official stack is too slow, too weak, or both. Ask your teams, plainly and without penalty, what they actually reach for and why. The honest answer tells you where your sanctioned path is failing them.
Pick one real process, not a platform
The firms seeing value did not roll AI across everything. They started with a specific piece of work, designed it well, and proved the return. Choose one process where quality, not just speed, would change your client relationship. Map how the work flows today. Decide where judgment has to stay human. Then build the AI to fit that design.
Set the quality bar before the speed target
Hasker's point about professional outputs applies to your firm, not just to your vendors. Decide what good enough to deliver means for the work in question, and put a review step there on purpose. Quality you can defend is what 78% of clients are asking for. It is also what keeps you off the reconsideration list.
Treat AI access as a retention and recruiting question
If 62% of professionals weigh AI access when they choose a job, then your tooling is part of your offer. The people most likely to leave over this are usually the ones doing the best work. Giving them professional-grade tools inside a well-designed process is a retention decision, not an IT line item.
The next step
Pick one process this week. Not a platform, not a strategy deck. One piece of work where doing it better, not just faster, would change a client relationship or free up your best people. Map how it runs today and mark the one place a human has to stay in the loop. That single exercise tells you more about your real AI readiness than any adoption dashboard, and it is the first step the firms in this report took before the value showed up.
If you want a structured read on where your organization sits today, The 7 Levels of AI Proficiency assessment gives you a baseline in about ten minutes. It is a starting point for the same question this report is asking: not whether your people use AI, but whether your organization is built to get real value from it.
Related reading: Level 5: The Captain (Design Thinker).
Sources
- AI is Ready but Firms are Not: How Falling Behind on AI Implementation is Costing Clients and Talent (Thomson Reuters)
- AI is Ready but Firms are Not (Future of Professionals Report 2026 wire release, PR Newswire)
Frequently Asked Questions
Who did this report survey?
Thomson Reuters surveyed 1,816 professionals across law, tax, audit, accounting, compliance, risk, and global trade, in March and April 2026, spanning private-practice firms and in-house corporate and government departments across 62 countries.
Is the $143 billion figure independently verified?
No. It is Thomson Reuters' own estimate, modeled from the survey: 32% of clients reconsidering provider relationships within 12 months, with a third of those putting more than $1 million in annual work at risk, totaled across U.S. legal and accounting revenue. It is sourced and attributable to Thomson Reuters, not an outside accounting of confirmed losses. Useful as direction, not as a precise ledger.
My team already uses AI every day. Aren't we fine?
Daily usage is the starting point, and 74% of professionals are already there. The report's central finding is that usage and value are not the same thing. 91% of those daily users still believe their organization falls short of AI's potential. The return comes from how the work is designed, not from how often the tools get opened.
What does "operationalizing AI" actually mean here?
Building AI into a designed process with a defined quality bar and a clear place for human judgment, instead of leaving people to improvise with whatever tool is handy. The report ties that practice to the firms pulling ahead on both client and talent retention.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.