The Legal AI 'Deposition Prep' Asymmetry: Why Plaintiffs' Firms Are Using AI to Cross-Reference Witness Histories While Defense Teams Are Still Doing It Manually
The deposition room has always rewarded preparation. In 2026, preparation increasingly means something specific: running a witness's prior sworn testimony, public statements, regulatory filings, and social media history through an AI system trained to surface contradictions before the examination begins. Plaintiffs' firms — particularly boutiques...
The deposition room has always rewarded preparation. In 2026, preparation increasingly means something specific: running a witness's prior sworn testimony, public statements, regulatory filings, and social media history through an AI system trained to surface contradictions before the examination begins. Plaintiffs' firms — particularly boutiques operating in mass tort, employment discrimination, and consumer class actions — are doing this routinely. Many BigLaw defense teams are still handing the job to a third-year associate with a legal pad and a Relativity login.
This is not a subtle gap. It is a structural asymmetry with real consequences in deposition rooms, and the legal industry's collective reluctance to call it a crisis says more about billing incentives than it does about capability.
How Plaintiffs' Firms Are Actually Doing This
The tooling has matured considerably since the early promise of 2023. Firms running contingency-fee mass tort litigation — think personal injury boutiques handling PFAS cases, talc litigation, or employment discrimination class actions under Title VII — are now deploying purpose-built deposition intelligence platforms alongside general-purpose legal AI infrastructure.
Tools like CaseText's CoCounsel, Relativity aiR for Review, and emerging purpose-built platforms such as Deposiq and EvenUp (primarily settlement-focused but increasingly used upstream) allow attorneys to ingest a witness's entire deposition history — across cases, jurisdictions, and years — and prompt the system to identify inconsistencies between what the witness said in a 2019 product liability case and what they are expected to testify in an examination scheduled for next Tuesday. Combine this with public records aggregation — SEC filings, OSHA violation records, FDA adverse event reports, LinkedIn activity, and archived press statements — and you get a witness profile that a single associate working a 60-hour week simply cannot replicate with comparable accuracy or completeness.
Morgan Lewis partner and repeat expert witness Dr. Patricia Dunmore, for example, might have testified in 14 different pharmaceutical cases since 2017. An AI system can cross-reference those transcripts in hours. A human can do it in a week, if they catch every instance and if the firm bills for the time.
That last clause is the point.
The Incentive Structure Explains Everything
The asymmetry exists because of a fundamental misalignment between how plaintiffs' firms and defense firms are paid. Plaintiffs' boutiques operate on contingency. Every dollar spent on associate hours reviewing deposition transcripts manually is a dollar that reduces the firm's recovery. The incentive to automate is direct, immediate, and financial. If AI can surface a witness's contradictory testimony from a 2021 EEOC proceeding in 45 minutes instead of 14 hours of associate billing, the boutique firm captures that efficiency as profit and competitive advantage.
BigLaw defense teams operate on hourly billing. Associate review time is not a cost — it is revenue. A partner overseeing a Fortune 500 client's employment discrimination defense has no structural incentive to replace 40 hours of billable associate work with a $500 AI query, even if the AI query is demonstrably more thorough. The client might prefer it. The firm's revenue model does not.
This is not a technology problem. It is a billing model problem dressed up as a capability gap.
How Defense-Side Partners Are Rationalizing the Lag
The rationalizations are predictable and, frankly, unconvincing. The most common is the quality control argument: that AI systems hallucinate, miss context, and require attorney review anyway, which means the time savings are illusory. There is a grain of truth here — unverified AI outputs in deposition prep are genuinely dangerous — but this argument proves too much. The plaintiffs' boutiques using these tools are not skipping attorney review. They are using AI to generate the roadmap that attorneys then evaluate and refine. The review is faster and better because the system has already done the pattern recognition.
The second rationalization is client conflict: that sophisticated corporate clients, particularly in regulated industries, have data privacy concerns about uploading sensitive case materials to third-party AI platforms. This is a real and legitimate issue in some contexts, particularly for matters touching on HIPAA-regulated data or export-controlled information. But it does not explain the broader lag, and most enterprise AI platforms in legal have built data governance structures sufficient to address standard confidentiality concerns under Model Rule 1.6.
The third rationalization is the most honest and the least spoken: institutional inertia. BigLaw partnership structures are not optimized for rapid technology adoption. The partners who built their practices on associate leverage are not rushing to automate it away.
What This Looks Like in the Room
The practical consequences are visible in 2026 deposition rooms in specific, documentable ways. Plaintiffs' counsel in employment discrimination cases are arriving with precision cross-examination outlines built on years of a corporate HR director's prior sworn statements in EEOC proceedings, arbitration transcripts, and internal investigation interviews produced in discovery. Defense counsel, working from manually assembled witness prep binders, are encountering impeachment sequences they did not anticipate because they did not have the full picture of their own witness's testimony history.
In mass tort multi-district litigation — where corporate witnesses testify repeatedly across hundreds of individual cases — this asymmetry is most acute. The plaintiffs' MDL coordinating committee has a comprehensive AI-generated witness history. The defense team is managing their fourth set of local counsel handoffs and hoping someone checked the 2020 transcript.
This Is a Billing Model Problem, and the Clock Is Running
Calling this a capability crisis lets the defense bar off the hook. The capability exists. The decision not to deploy it is structural and largely financial. The more honest framing is that BigLaw's hourly billing model has created a perverse incentive against adopting technology that would benefit clients but reduce firm revenue — and that clients are beginning to notice.
General counsel offices that have watched plaintiffs' counsel run circles around defense teams in deposition preparation are not going to be patient indefinitely. Some sophisticated clients are already demanding AI-assisted deposition prep as a line item in litigation budgets, forcing the conversation. When clients lead technology adoption because their outside counsel won't, the billing model has become indefensible.
The deposition room is where preparation becomes outcome. Plaintiffs' firms figured that out first.