The AI Trust Gap Law Firms Can No Longer Ignore

The uncomfortable question about artificial intelligence in the legal profession is no longer whether lawyers are experimenting. Many are. The harder question is whether the client paying for the matter knows where AI touches the work, who reviews the output, whether the bill changed, and whether anyone measured the result.

That is where the relationship is starting to strain. Thomson Reuters reported in 2026 that 68% of corporate legal professionals do not know whether their outside counsel use AI on their matters; 60% do not know whether those firms use generative AI specifically.[1] The numbers matter less as a technology snapshot than as a governance failure. A legal department cannot assess value, risk, staffing, or supervision if the firm relationship depends on guessing.

The source of the 68% figure is worth naming plainly. Thomson Reuters is not a neutral academic observer of the legal AI market; it sells legal technology, including AI-enabled products. But the finding should not be dismissed as vendor enthusiasm. It lines up with a broader pattern: individual lawyers are adopting AI faster than institutions are explaining it, while clients are expecting benefits faster than they are building disciplined questions into outside counsel management.

Adoption Is Moving Faster Than Disclosure

The usual adoption story is too simple. It suggests a legal market divided between forward-looking AI users and laggards. The more useful distinction is between individual use, institutional readiness, and client-visible governance.

On the in-house side, the pace has already changed. Citing ACC and Everlaw survey data, Thomson Reuters reported that in-house generative AI adoption rose from 23% to 52% in one year, and that 64% of in-house teams expected to depend less on outside counsel because of internal AI capabilities.[2] Those figures are from 2025, not a real-time 2026 census, but they are recent enough to explain why general counsel and legal operations teams are now bringing different assumptions into budget conversations.

Inside law firms, adoption also appears substantial, though not always institutionalized. LawNext’s coverage of 8am’s 2026 report said 69% of legal professionals reported personal AI adoption, while firms lagged far behind individual practitioners in formal implementation.[3] That distinction is critical. A senior associate using AI to accelerate a first-pass chronology is not the same thing as a firmwide workflow with approved tools, client-data rules, review standards, pricing treatment, and matter-level reporting.

Clients often experience that distinction as silence. They may suspect AI is being used because turnaround times have shifted, drafts look different, or the firm has mentioned AI in a pitch. But suspicion is not supervision. It does not tell the client whether the firm used an enterprise tool, a public tool, a practice-group pilot, or no AI at all. It does not tell the client whether confidential data entered the system. It does not tell the client whether a partner reviewed the output differently from any other work product.

That last row is where in-house teams should feel a little less comfortable. Legal departments can complain that firms are opaque, and often they are right. But many clients have not made AI disclosure a normal part of engagement letters, billing guidelines, panel reviews, matter kickoffs, or quarterly business reviews. If the only AI question comes after a surprising invoice or a late-stage quality issue, the conversation is already defensive.

The Expectation Gap Is Now Commercial

The trust problem would be easier to contain if clients were merely curious. They are not. Thomson Reuters reported that 78% of corporate clients said AI-enabled quality improvements were essential or very important, while only 6% said most of their providers were delivering those improvements.[4] That is not a small service-delivery gap. It is the kind of mismatch that eventually shows up in panel reviews, RFP scoring, matter allocation, and fee discussions.

The sharper point is that clients are not only disappointed in an abstract way. Thomson Reuters also reported that 32% of corporate clients are reconsidering relationships with law firms they view as falling behind on AI.[1] That does not mean one-third of clients are about to replace firms over a chatbot policy. It means AI has entered the relationship-risk file. For a relationship partner, that should be enough to bring the issue out of the innovation committee and into client management.

Still, expressed expectations are not the same as active buying behavior. The 8am report, as covered by LawNext, found that only 6% said clients were explicitly pushing for AI-linked cost reductions.[3] That figure is irritating because it exposes a passive form of dissatisfaction. Clients want better quality and may expect efficiency, but relatively few are putting a clear pricing demand on the table.

There are understandable reasons for that hesitation. Some legal departments do not yet know which AI uses they want to encourage, prohibit, or price differently. Some worry that asking too broadly will create a burdensome disclosure exercise with little practical value. Others do not want to reward firms for reducing hours if the underlying fee arrangement still treats hours as the main measure of value. Those are real concerns, but they do not justify leaving the subject to hallway comments and conference-panel optimism.

The Billing Anxiety Nobody Can Ignore

The sentence that will stay with many clients is the one Thomson Reuters quoted from a chief legal officer: “I fear firms will use AI to cut time but continue billing for hypothetical time.”[1] It should be handled carefully. The quote is not proof that firms are padding bills or systematically hiding AI-driven efficiencies. It is evidence that some clients can already imagine the abuse because the current conversation gives them too little else to work with.

That fear lands in a market where many matters are still budgeted, staffed, and reviewed through time. If AI reduces the associate hours needed for a first draft, privilege log, contract comparison, deposition summary, or research memo, someone has to decide what happens next. Does the firm write down time? Does it keep the same fixed fee because the value was the answer, not the labor path? Does it redeploy senior review time to improve quality? Does the client receive a different staffing model next time?

There is no universal answer, and pretending there is would flatten too many fee models. But the absence of a universal answer is exactly why the issue belongs in matter planning. A firm does not need to disclose every keystroke to have a serious conversation about where AI may affect scope, staffing, timing, quality control, and price. A client does not need perfect internal policy maturity to ask whether AI use is built into the proposed budget.

Both Sides Say the Firm Should Start. Then Many Wait.

One of the more revealing Thomson Reuters findings is not about tools at all. Seventy-five percent of both law firm and corporate respondents said it is the law firm’s responsibility to initiate AI discussions.[1] That shared expectation should make the next step easy. It apparently has not.

Firms may be reluctant for several reasons. They may worry that disclosure will trigger fee pressure before they have a pricing strategy. They may not have a clean inventory of AI use across practice groups. They may be concerned that one client’s preferred disclosure standard will conflict with another’s. They may also be trying to avoid saying the quiet part out loud: individual lawyers are moving faster than firm governance.

Clients have their own avoidance patterns. A legal department may say it wants innovation but keep outside counsel guidelines unchanged. It may ask about AI in an RFP but never revisit the answer during active matters. It may demand efficiency from firms while declining to define whether savings should appear as lower fees, faster cycle time, better work product, reduced internal review burden, or some combination of those outcomes.

That mutual waiting creates a bad operating environment. Firms infer that clients are not serious because few explicitly demand AI-linked cost reductions. Clients infer that firms are hiding something because firms do not volunteer enough detail. Neither inference is reliable. Both can damage the relationship.

Measurement Is the Missing Control

The trust gap would be less dangerous if either side had strong measurement. Thomson Reuters reported that 85% of law firms and 75% of corporate legal departments were either not collecting AI ROI data or were unsure whether they were collecting it.[1] In practical terms, that means many organizations are discussing AI value without a reliable baseline.

ROI does not have to start as a grand dashboard. At matter level, the first useful measures are usually plain: cycle time, hours by phase, write-offs, budget variance, number of review rounds, internal client satisfaction, error rates where trackable, and whether senior lawyers spent more time on judgment rather than assembly. A firm can capture some of this without turning every matter into a lab experiment. A legal department can ask for some of it without demanding privileged work-product detail.

The most important move is to separate adoption evidence from value evidence. “Our lawyers use AI” is adoption evidence. “This workflow reduced first-draft turnaround while preserving partner review and lowering the agreed fee for repeat work” is closer to value evidence. “We have an AI committee” is governance evidence, but it is not yet a client outcome.

What Firms Should Put on the Table

Law firms do not need to turn every client conversation into a product demo. Most clients do not want that. They want to know whether the firm can protect the matter, supervise the work, and explain the value. A useful AI discussion is therefore closer to matter management than marketing.

• Describe where AI may appear in the workflow: legal research, document review, contract analysis, drafting support, summarization, knowledge retrieval, or administrative work.
• State which uses require client discussion before deployment, especially where client confidential information, sensitive data, high-risk advice, or unusual tool settings are involved.
• Identify the human review model: who checks the output, at what stage, and whether review responsibility changes by task risk.
• Explain the fee treatment: whether AI-enabled efficiencies affect hourly estimates, fixed-fee assumptions, caps, success fees, or future repeat-work pricing.
• Offer basic measurement: before-and-after cycle time, phase-level staffing changes, budget performance, quality indicators, or lessons learned from similar matters.

The disclosure should be specific enough to matter and restrained enough to be repeatable. A client does not need a 20-page technical appendix for every routine matter. But “we use AI where appropriate” is now too vague to carry trust. It answers none of the questions that matter to the person approving the budget.

What In-House Teams Need to Ask Before the Invoice Arrives

Clients also have work to do. Waiting for every firm to volunteer a perfect AI disclosure protocol is not a management strategy. If AI matters to quality, risk, cost, or timing, it belongs in the same places where clients already manage outside counsel: engagement letters, billing guidelines, RFPs, matter kickoffs, status calls, and business reviews.

• Which AI uses are permitted without advance notice, which require notice, and which require explicit approval?
• May client confidential information be entered into AI tools, and under what security, retention, and training restrictions?
• How will the firm supervise AI-assisted work, and who remains accountable for the final advice?
• How should AI-enabled efficiencies affect budgets, staffing assumptions, alternative fee arrangements, or future matter estimates?
• What simple performance data will the firm provide at matter close or during the next business review?

Those questions should not be reserved for firms perceived as technologically advanced. The greater risk may be the middle of the panel: firms with enthusiastic individual lawyers, uneven practice-group controls, and no standard client-facing explanation. That is where a legal department can believe it has hired a sophisticated provider while still lacking visibility into the actual work path.

The Firms Most Exposed Are the Ones Letting Clients Guess

By Q3 2026, AI use in legal work is no longer novel enough to be handled as an occasional innovation update. It affects how clients think about quality, cost, speed, staffing, and dependence on outside counsel. It also affects how firms defend their value when clients are building more internal capability.

The firms most exposed are not necessarily the ones using too little AI. A firm can be cautious for good reasons, especially in high-risk work. The more dangerous position is using AI, marketing AI, or benefiting from AI-enabled efficiency while leaving clients to infer what changed. Silence may avoid one awkward pricing conversation this quarter. It can also train the client to believe the firm is managing the most important service-delivery change in years without a shared record of the rules.

That is the trust gap law firms can no longer ignore: whether the people responsible for the matter can explain how AI changes the work, the bill, and the basis for confidence.

References

1. The great AI disconnect: Law firms & legal departments are not communicating about AI usage — Thomson Reuters
2. See what legal professionals say about the role of AI and law — Thomson Reuters
3. AI Adoption Among Legal Professionals Has More Than Doubled in a Year, New 8am Report Finds — LawNext
4. 2026 Future of Professionals Report — Thomson Reuters