The technology is powerful. The judgment about how to use it — and when not to — still belongs to the lawyer.
The courtroom is the last place you’d expect a tech revolution: judges still wear robes; lawyers still cite cases from the nineteenth century; oral argument still looks, for the most part, exactly as it did a hundred years ago. Litigation is one of the most tradition-bound professions in existence, a field where the written brief and the spoken word remain the fundamental currency of advocacy.
And yet, quietly and faster than most people realize, artificial intelligence is restructuring how lawsuits are built, fought, and won. According to recent analysis, 44% of legal work tasks could be automated by generative AI – the second-highest automation rate of any industry in the United States.
In this article, I will shed some light on what is changing right now: in how cases are prepared, how documents are reviewed, how strategies are built, and how the legal system itself is beginning to grapple with tools it was never designed to handle. Some of these changes are genuinely good for clients and for the justice system. Some carry risks the profession has not yet fully reckoned with. But all of them are worth understanding, whether you are a lawyer, an investor, a business owner, or anyone who has ever been involved in a legal dispute.
Part I: What AI Is Actually Doing in Litigation Today
Document Review and Discovery
The most immediate and widespread application of AI in litigation is document review. In complex commercial cases, such as securities fraud class actions, partnership disputes, and antitrust matters, the discovery process can involve millions of emails, contracts, financial records, and internal communications. Historically, reviewing those documents required armies of junior associates working through the night, billing by the hour, and introducing the kind of human inconsistency that inevitably comes with fatigue and volume.
Technology-assisted review (TAR) and predictive coding changed that equation significantly, and generative AI is accelerating the shift further. These tools use machine learning to identify which documents are likely to be relevant, allowing legal teams to focus human attention where it matters most. The results are striking: firms using automated e-discovery workflows reduce review costs by 60% and processing time by 70% compared to traditional linear review. In large-scale litigation, that translates directly into lower legal spend and faster resolution.
For the clients footing the bill, this matters enormously. Discovery has historically been the single largest driver of litigation cost. Anything that reduces the volume of hours required for document review has a direct and significant impact on what litigation actually costs.
Legal Research
AI has also transformed how lawyers research the law. Tools now surface relevant case law, flag circuit splits, identify how specific judges have ruled on specific arguments, and expose potential weaknesses in a legal theory before a brief is ever filed. Generative AI legal research now achieves approximately 92% precision per 2024 benchmarks — a level of accuracy that would have seemed implausible for an automated system only a decade ago.
A well-prepared brief used to require days of manual research. AI compresses that timeline substantially, and it does something manual research cannot do at all: it processes the entire universe of potentially relevant precedent simultaneously, without fatigue and without the cognitive biases that lead human researchers to anchor on the first strong case they find.
Drafting, Deposition Prep, and Trial Strategy
Generative AI is also being used for drafting — complaints, motions, discovery requests, contract clauses — and for deposition and trial preparation. Tools can now analyze thousands of pages of testimony to surface inconsistencies, flag credibility risks, and identify patterns across a witness’s prior statements that a human reviewer might miss under time pressure.
It is worth noting, however, that AI doesn’t produce a finished product. It accelerates the mechanics, but AI does not replace the legal strategy. The judgment required to craft a compelling legal argument — anticipating how a particular judge will read a brief, calibrating tone, structuring a narrative that will hold together under scrutiny — remains the work of an experienced attorney.
Where Adoption Stands Today
According to the American Bar Association’s 2025 Legal Industry Report, 31% of legal professionals now use generative AI personally for work-related tasks, up from 27% the prior year. Firm-wide adoption lags behind at just 21%, largely due to policy hesitation, data privacy concerns, and the significant cost of enterprise-grade legal AI systems. Civil litigation leads all practice areas in firm-level AI adoption at 27%.
The gap between individual use and firm-wide deployment shows that lawyers are adopting AI faster than their institutions can build policies around it.
Part II: The Strategic Shift
The End of the Discovery Arms Race
For decades, one of the most reliable litigation tactics available to a well-resourced party was document volume. Bury the other side in paper. Force them to review millions of pages with a team half your size and a budget a fraction of yours. The strategy worked because it was expensive to counter and courts were reluctant to police it aggressively.
AI has disrupted that calculus. When a smaller firm with the right tools can process what once required a large litigation department, the discovery arms race becomes less decisive. Preparation and strategy matter more.
Pattern Recognition at Scale
Perhaps the most consequential capability AI brings to litigation is pattern recognition across enormous datasets. The one email in four million that contradicts a witness’s sworn testimony. The financial record that shows a pattern of transactions inconsistent with what was disclosed to investors. The internal communication that establishes knowledge at the executive level three months before a public statement that said the opposite.
Human reviewers find some of these. AI finds more of them, faster, and without the confirmation bias that leads people to stop looking once they have found something useful. This changes settlement calculus. Parties who once felt safe in litigation because the key documents were buried deep in a large production can no longer rely on that safety.
Predictive Analytics and the Front-Loading of Strategy
AI tools now model litigation outcomes based on judge, jurisdiction, case type, and opposing counsel. According to recent benchmarks, predictive analytics can help forecast case outcomes with approximately 85% accuracy. While this is not a guarantee, it serves as a meaningful input into strategic decisions about whether to file, whether to settle, and what arguments to lead with.
The deeper strategic implication is about timing. Because AI can surface key facts and assess legal exposure earlier in a case, the center of gravity in litigation is shifting toward the beginning. Early case assessment — understanding what the documents actually show before the first brief is filed — is becoming more decisive. Lawyers who invest in that front-end work are better positioned than those who rely on the traditional approach of learning the case as it unfolds.
What Happens to the Billable Hour?
There is a tension at the heart of AI adoption in law that the profession has not yet resolved. If AI lets a lawyer accomplish in one hour what used to take five, the time-based invoice would shrink by 80% — despite the output being identical or better. Among firms that have widely adopted AI, 45% have already adjusted their pricing as a result.
Some firms have raised prices, arguing that faster, higher-quality work commands a premium. Some have reduced prices, passing efficiency gains to clients. Some have added AI-specific fees. The market has not settled on a model. For clients who have historically been priced out of sophisticated legal representation, this moment represents a genuine opportunity — if the profession chooses to pass savings through rather than absorbing them.
Part III: Risks
Hallucination and the Duty of Competence
In June 2023, attorneys in the Southern District of New York filed a brief in Mata v. Avianca that cited six court cases as legal authority. The cases did not exist. They had been generated by ChatGPT, complete with realistic-sounding case names, docket numbers, and fabricated judicial opinions. When the court demanded verification, the attorneys asked ChatGPT whether the cases were real. The AI confirmed that they were.
The court sanctioned the attorneys. But this was not an isolated incident. A researcher tracking AI hallucination cases in courts worldwide has now identified over 1,400 such cases, with the pace accelerating to roughly two new cases per week and growing. Courts have responded with increasing severity: in Johnson v. Dunn (N.D. Ala. 2025), a federal judge found monetary fines insufficient to deter the conduct and instead disqualified the offending attorneys from the case entirely, directing the clerk to notify bar regulators in every state where those attorneys are licensed.
Large language models generate probabilistic text. They do not know when they are wrong. They cannot distinguish a real case from a plausible-sounding fabrication. Attorneys who use these tools should always conduct an independent verification of AI-generated results.
The American Bar Association addressed this in Formal Opinion 512 (2024), confirming that the duties of competence, candor toward the tribunal, and supervision of nonlawyer assistance all apply to AI-assisted legal work. As of 2025, New York State now requires at least two annual CLE credits in AI competency.
Confidentiality and the Privilege Question
When a lawyer feeds client documents into a third-party AI platform to accelerate review, the natural question that arises is where that data goes. Large firms have addressed these concerns about data privacy with in-house systems or enterprise platforms with robust data protection agreements. But many smaller firms have not.
Client documents shared with a general-purpose AI tool without adequate data handling safeguards may not remain confidential. The rule of thumb is: if you would not hand those documents to an unvetted third party, you should not upload them to an unvetted AI platform.
The New Kind of Inequality
AI is not distributing its benefits evenly. Firms with 51 or more attorneys are using AI at roughly double the rate of smaller firms. The price tag of enterprise-grade AI systems is a significant barrier for smaller practices.
This matters not just for the economics of legal practice but for access to justice. Clients who are already disadvantaged by their inability to afford large-firm rates are now contending with a legal landscape in which their attorneys may be operating with materially inferior tools.
Bias, Precedent, and the Limits of Historical Data
AI systems trained on historical legal data will reflect historical outcomes. The American legal system has documented disparities in how cases involving race, class, and geography have been adjudicated. An AI tool that optimizes for precedent — recommending strategies likely to succeed based on how similar cases have been decided — is not recommending justice. It is recommending the historical mean.
The most important cases in legal history were precisely the ones that did not match the historical pattern. As predictive tools become more integrated into litigation strategy, there is a real risk that attorneys relying on AI-generated probability estimates may give up on potentially precedent-setting cases.
The Erosion of Judgment
The subtler risk, and perhaps the most consequential for the long run, is what happens to legal judgment when AI handles more and more of the analytical work. The ability to read a document carefully, to hold a complex factual record in mind, to sense when something in a witness’s account does not add up — these are skills developed through practice. They are not separable from the work that produces them.
Attorneys who use AI as a crutch for tasks that build those skills may find, over time, that the skills do not develop. The profession needs to think carefully about which uses of AI accelerate legal competence and which ones quietly erode it.
Part IV: What This Means for Clients
The potential client benefit from AI in litigation is real. The legal industry could save approximately $20 billion annually through automation of routine tasks. And some of those savings are already reaching clients in the form of reduced discovery costs and more competitive pricing from smaller firms that have adopted AI effectively.
But the distribution of those benefits is uneven. The clients most likely to benefit are those already well-served: sophisticated institutional litigants with legal teams that can evaluate AI tools, negotiate appropriate data-handling agreements, and verify AI outputs before they reach a court. Individual litigants and small business owners navigating a first serious commercial dispute face a more complicated picture.
For those clients, the most important thing an attorney can do has not changed: understand the facts of your matter better than anyone else in the room, prepare more thoroughly than the other side expects, and exercise the kind of judgment that no algorithm currently replicates. AI can support all of those things. It cannot substitute for any of them.
The practical advice for anyone involved in or considering litigation is straightforward: ask your attorney how they are using AI tools, what safeguards are in place for your confidential information, and how they are verifying AI-generated work product before it goes into a court filing. Those are fair questions, and any attorney worth hiring should be able to answer them clearly.
Part V: Where This Is Headed — The Next Five Years
The trajectory of the legal AI market gives some sense of the scale of what is coming. The global legal AI market was valued at $1.45 billion in 2024 and is projected to reach $3.90 billion by 2030, growing at 17.3% annually.
AI-assisted judicial administration is already being piloted in some jurisdictions for case management, scheduling, and law clerk research. The line between administrative and substantive judicial functions is not always clean, and courts will need to develop clear policies about where AI assistance ends and judicial judgment begins.
AI-generated evidence is an emerging challenge courts are only beginning to address. Deepfakes, synthetic documents, and AI-enhanced audio and video create evidentiary questions that existing authentication frameworks were not designed to answer. Litigators will need to develop expertise in identifying and challenging this kind of evidence.
Real-time AI assistance during proceedings — tools that monitor testimony for inconsistencies, surface relevant documents mid-examination, or flag legal issues as arguments are made — already exists in early commercial form. As these tools mature, the rules governing what is permissible will need to keep pace.
Perhaps most striking is the profession’s own assessment: 80% of legal professionals expect AI to have a transformative or high impact on their work over the next five years, and 95% believe AI will be a central component of their workflows within that window. Among those who have already adopted AI widely, 69% report a positive impact on revenues.
Conclusion: The Profession Has Adapted Before — But This Time Is Different
The legal profession has absorbed technological disruption before. The photocopier changed how documents were reproduced. Electronic filing changed how cases moved through courts. E-discovery changed how evidence was collected and reviewed. Each of these transitions was disruptive in its moment and is now simply part of how the practice works.
AI is different in degree and in kind. It is not changing how existing tasks are performed. It is changing which tasks require human beings at all. The distinction matters because the skills that make a great litigator — judgment, strategy, the ability to read people and situations, the capacity to construct a persuasive narrative under pressure — are precisely the things AI cannot replicate. They are also the things most at risk of being neglected when AI handles everything else.
The lawyers who will thrive in this environment are not the ones who resist the technology. They are the lawyers who understand what AI does well enough to use it effectively, who maintain the human judgment that AI cannot substitute for, and who think clearly about the ethical obligations that attach to both.
Chief Justice John Roberts noted in his 2023 year-end report on the federal judiciary that any use of AI “requires caution and humility.” That is as good a summary of the professional obligation as any. The technology is powerful. The judgment about how to use it — and when not to — still belongs to the lawyer.
Join the conversation!