What corporate clients demand from law firms on AI in 2026

The question corporate clients are asking about AI in law firms is no longer whether the firm has heard of generative AI. It is whether the firm can explain, matter by matter, what AI is doing, what a lawyer checks, what happens to confidential information, and whether the client is being billed as if none of the efficiency exists.

That is why the most useful 2026 data point is not an adoption statistic. It is the gap between expectation and delivery: 78% of corporate clients say AI-enabled quality improvements are very important or essential, while only 6% say most of their providers deliver those improvements. In the same Thomson Reuters report, 32% of corporate clients say they are reconsidering relationships with firms they perceive as falling behind on AI.[1]

For a general counsel, that gap is not abstract. If a board member, CFO, audit committee, or CISO asks how outside counsel is using AI on sensitive work, “they use it responsibly” is not an answer. It does not identify the workflow, the data exposure, the review point, the quality control, or the pricing consequence. It leaves the legal department holding accountability without enough information to supervise the work.

The client is not just waiting for the firm to catch up

The commercial risk for firms is sharper because in-house teams are not standing still. In an ACC/Everlaw survey, 64% of in-house legal teams said they expect to depend less on outside counsel because of AI capabilities they are building internally. The same survey reported that corporate legal department AI adoption doubled from 23% to 52% in one year.[2]

That 64% figure should be read carefully. It is an expectation, not proof that work has already moved or that every legal department can replace outside counsel on complex matters. Some departments will overestimate what their tools, data, staffing, and risk controls can support. But as a buying signal, it matters. Clients are telling the market that the default flow of work to outside counsel is becoming more contestable.

The pressure is not simply “do it cheaper.” Corporate clients are asking for a hard combination: better quality, faster execution, clearer controls, and some recognition that AI may reduce effort in certain parts of the work. Those goals can conflict. A firm that uses AI to accelerate first-pass review may still need senior lawyer time to validate the answer, handle judgment calls, and protect privilege. The point is not that every AI-assisted task should become cheap. The point is that the firm should be able to explain where the work changed and why the fee still makes sense.

What clients mean when they ask for AI value

The phrase “AI-enabled quality” can sound vague until clients translate it into review criteria. In Thomson Reuters’ Future of Professionals data, 96% of respondents demanded safeguards for confidential data, 94% demanded outputs grounded in authoritative content, and 90% said AI must produce explainable reasoning.[3]

Those expectations are practical. A legal department does not need a tour of the firm’s innovation lab to approve a workflow. It needs to know whether a contract summary tool is operating inside an approved environment, whether privileged material is excluded from model training, whether citations are checked against primary or otherwise authoritative sources, and whether the final advice reflects lawyer judgment rather than a polished machine-generated draft.

The confidentiality point deserves more weight than it often gets in marketing conversations. Many corporate legal teams now sit inside broader enterprise risk systems. They answer to procurement questionnaires, security reviews, data retention policies, incident response plans, and sometimes regulatory obligations. If outside counsel cannot say where client data goes, the legal department may have to slow or block the workflow even if the legal team likes the result.

Grounding is the second test because legal work punishes confident unsupported answers. A firm may use AI for issue spotting, summarization, deposition preparation, diligence review, or research acceleration. Those uses are not equivalent. A summary of documents already in the matter file raises different questions than a generated statement of law. Clients increasingly want to know which sources the system can draw from and how the firm prevents a fluent answer from outrunning the authority behind it.

Explainability is where supervision becomes visible. The client does not need every input. It does need confidence that the lawyer reviewing the output can reconstruct why the answer was given, identify what was excluded, and take responsibility for the final work product. If the lawyer cannot explain the AI-assisted conclusion, the client has no reason to treat it as improved quality.

The communication channel is broken

Law firms are not operating in a clean instruction environment. Thomson Reuters reported that 40% of firm respondents had received conflicting client instructions telling them both to use and not use AI. The same source found that fewer than one-third of corporate departments know whether their outside firms use AI on client matters.[1]

That combination explains a lot of the current dysfunction. Some clients want AI-driven efficiency but prohibit unapproved tools. Some want disclosure but have not decided what level of disclosure they mean. Some business stakeholders want legal spend reduced, while the legal, privacy, and security teams are still debating acceptable use. A relationship partner who waits for one clean instruction may be waiting a long time.

Still, the uncertainty does not excuse silence. If a firm receives inconsistent positions across its client base, that is exactly the reason to build a matter-opening conversation. The firm can ask whether the client permits AI use, identify prohibited categories of data, confirm approved tools or environments, specify when human review occurs, and document any client-specific restrictions. That is not a burden added to the relationship; it is part of making the relationship supervisable.

The firms that handle this well do not need to overclaim. A modest workflow can be more credible than a sweeping promise. For example, a firm might explain that it uses an approved AI tool to generate first-pass chronologies from matter documents, that no client documents are used to train the model, that associates validate source references, and that the responsible partner reviews the final chronology before it is used in advice. That kind of description gives a client something to assess.

By contrast, “we have an AI policy” is not enough. Most clients assume a serious firm has policies. The harder question is whether the policy has been translated into engagement letters, outside counsel guidelines, billing narratives, staffing assumptions, knowledge management, vendor review, and matter-level controls.

Adoption is now background, not differentiation

Broad legal AI adoption data makes the silence harder to defend. The 8am 2026 Legal Industry Report, published via ABA Law Practice Magazine and based on more than 1,300 legal professionals, reported that 69% use generative AI for work. It also found that 42% use legal-specific AI tools, up from 21% in 2025.[4]

Those figures do not prove that every firm has mature AI governance or that every use improves client work. They do show that AI is no longer unusual enough to be treated as a side topic. If lawyers across the market are using these tools, corporate clients are reasonable to ask whether their own matters are affected.

The distinction matters because adoption and effectiveness are often blurred. A firm can have many users and still have weak controls. Another firm can have narrower deployment and stronger review. Clients should care less about the number of licenses and more about whether the firm can connect a particular use case to a defensible workflow.

Pricing is part of the AI conversation

Billing is where many AI conversations become uncomfortable, which is why it should be addressed early. If AI reduces time spent on a task that was historically billed by the hour, the client will eventually ask who captures the benefit. The answer does not have to be a universal discount. It does have to be coherent.

A firm may decide that AI allows it to offer fixed fees for specific workstreams, faster turnaround for the same fee, capped fees with clearer assumptions, or blended staffing that moves more lawyer time toward analysis and strategy. In some matters, AI may increase quality control costs because lawyers spend more time validating outputs or because the firm invests in secure infrastructure. Those are legitimate considerations. They are easier to defend when the firm has already explained what changed in the work.

The worst position is to advertise AI efficiency in a pitch and then submit bills that look exactly like the old process. That gap invites procurement scrutiny and erodes trust with the legal department, which may already be under pressure to show that outside counsel spend is being managed.

New entrants are exploiting the transparency gap

AI-native legal providers are not yet a proven replacement for traditional firms across the full range of corporate legal work. The evidence is still early, and much of the visibility comes from industry reporting rather than long public track records. But they are aiming directly at the parts of the client experience that frustrate corporate legal departments: unclear process, hourly billing, and limited transparency into how technology affects the work.

Lupl’s 2026 overview of AI law firms highlighted companies including Crosby, Garfield AI, Avantia, Eudia, and Manifest, describing AI-native models built around fixed fees, no-billable-hour structures, AI-enabled workflows, and client transparency. The same source reported significant venture funding for some entrants, including Eudia at $105 million and Manifest at a $750 million valuation after raising $60 million.[5]

That does not mean a heavily funded entrant can handle a bet-the-company investigation, a cross-border dispute, or a regulatory crisis better than an experienced law firm. It does mean traditional firms should not assume that brand trust alone will protect work that is repeatable, process-heavy, or poorly explained. When a competitor packages the workflow, fee, and AI controls in plain language, the incumbent firm’s vagueness becomes a business-development problem.

Ethics guidance supports discipline, not paralysis

Ethical caution is real. Lawyers have duties of competence, confidentiality, supervision, candor where applicable, and reasonable billing. ABA Formal Opinion 512 addresses generative AI in relation to professional responsibility, including competence, confidentiality, supervision, and fees, but it does not create a single disclosure script for every matter.[6]

State requirements and expectations can vary, and firms should be careful about turning one jurisdiction’s guidance into a national rule. But that variation is not a reason to avoid the client conversation. It is a reason to make the conversation specific: what tool, what task, what data, what review, what jurisdiction, what client instruction.

What differentiated firms should make legible

A corporate client does not need every internal detail of the firm’s technology stack. It does need enough information to decide whether the AI use is acceptable, valuable, and priced fairly. The strongest firms will make that information available before the client has to chase it.

• Initiate the AI discussion at matter intake, especially for matters involving sensitive data, large-scale review, recurring work, or alternative fee arrangements.
• Map AI use by workflow rather than by tool name: research, document review, contract analysis, chronology creation, drafting support, knowledge retrieval, or billing administration.
• State what data enters the system, whether it leaves the firm or client environment, whether it is retained, and whether it can be used for model training.
• Describe the lawyer review point: who checks the output, against what source material, and before which client-facing or court-facing use.
• Explain how AI affects staffing, timing, fees, and billing narratives, including when efficiency is reflected in a fixed fee, cap, discount, or changed scope.
• Invite client-specific instructions and preserve them in the engagement record, outside counsel guideline addendum, or matter plan.

This does not require every partner to become a technologist. It does require the firm to stop treating AI as either a marketing slide or a forbidden topic. The relationship partner should know enough to say where AI may enter the work, when the client’s approval is needed, and who inside the firm owns the controls.

For firms that already have credible internal systems, the opportunity is immediate. Many clients are not asking for perfection. They are asking for a usable explanation. In 2026, AI capability alone is not enough; the firms most likely to gain ground are the ones that make their AI use legible to corporate clients before clients have to ask.

References

1. Thomson Reuters 2026 AI in Professional Services Report, Thomson Reuters.
2. ACC/Everlaw GenAI Survey, ACC and Everlaw.
3. Thomson Reuters Future of Professionals Report, Thomson Reuters.
4. 8am 2026 Legal Industry Report, ABA Law Practice Magazine.
5. 10 AI Law Firms to Watch in 2026, Lupl, 2026.
6. Formal Opinion 512, American Bar Association.