Consider a hypothetical commercial dispute, Meridian Analytics, Inc. v. Halvorsen Data Partners, LLC, in which the plaintiff alleges that a former licensee used Meridian's proprietary datasets beyond the scope of their agreement and then built a competing product on them. Not long ago, the early months of that case would have been dominated by volume. A review team would code several hundred thousand emails and chat messages for responsiveness and privilege, and associates would research the elements of misappropriation under the governing state's trade secret statute and the federal Defend Trade Secrets Act. Much of that work can now be done faster, and in some respects more consistently, by software. What cannot be delegated is the set of decisions the work was meant to inform: which theory to lead with, whether to seek a preliminary injunction, what to concede, and when to settle. This post argues that AI is moving the value of legal work from the first category to the second, and that lawyers and firms who see only a change in tools will be poorly placed to compete.
AI is absorbing the routine processing that once filled associate hours
Litigation has always contained a large layer of information processing beneath its strategic decisions. Lawyers read documents to classify them, read transcripts to summarize them, and read cases to determine whether they say what a brief needs them to say. Technology-assisted review has been accepted by federal courts for more than a decade [1], and generative tools now extend automation to tasks that once seemed to require a lawyer's reading. These include summarizing deposition transcripts, building chronologies from produced documents, drafting first versions of meet-and-confer letters, and identifying candidate authorities for a research question. In Meridian, a review platform could prioritize the messages most likely to discuss the licensed datasets, and a drafting tool could produce a serviceable first set of Rule 34 requests in minutes.
The economic consequence follows directly. Work that is billed by the hour and can now be performed in a fraction of the time will be billed for less, either because clients refuse to pay for the old hours or because competing firms stop charging them. Corporate legal departments have already begun asking outside counsel how AI is used on their matters and what it saves [2]. Information processing is becoming a commodity input. A lawyer whose value rests mainly on the capacity to process information quickly is competing with software on software's terms, and that is a contest the lawyer will eventually lose.
Professional duties keep the lawyer accountable for every machine-assisted output
Automation of the processing layer does not move responsibility anywhere, and the rules of conduct and procedure are clear on the point. ABA Model Rule 1.1 requires competence, and Comment 8 extends that duty to understanding the benefits and risks of relevant technology, a comment most states have adopted in some form [citation needed]. The ABA addressed generative AI directly in Formal Opinion 512, issued in July 2024, which discusses competence, confidentiality under ABA Model Rule 1.6, client communication, supervision, and the reasonableness of fees [3]. At the procedural level, an attorney who signs a brief certifies under Fed. R. Civ. P. 11(b)(2) that its legal contentions are warranted by existing law or a nonfrivolous argument for changing it, and an attorney who signs discovery responses certifies under Rule 26(g) that they were made after a reasonable inquiry [4][1]. None of these obligations is discharged by pointing to the tool that produced the draft.
Courts have made the consequences concrete. Lawyers in several jurisdictions have been sanctioned for filing briefs that cited cases a generative tool invented [4][5], and a number of federal judges have issued standing orders requiring disclosure or certification of AI use in filings [citation needed]. These episodes are usually told as cautionary tales about careless individuals, but they carry a broader lesson about where value now lies. Checking authority against its source, and judging what a document does and does not support, has become the lawyer's distinctive contribution to a machine-assisted draft. A lawyer who can do that reliably and quickly is worth more than before, while one who cannot is a liability to the client and the firm.
Strategic judgment becomes the scarce input in litigation
When processing is cheap, the scarce resource is the capacity to decide what the processed information means for the case. Suppose the review in Meridian surfaces messages in which a Halvorsen engineer refers to "the Meridian tables" months after the license expired. A summarization tool can find those messages and place them on a chronology. It cannot tell counsel whether to use them now in a motion for a preliminary injunction, which requires among other things a showing of likely success on the merits and irreparable harm [6], or to hold them for deposition, where the engineer's explanation can be tested before the defense has shaped it. That choice turns on the client's commercial position, the court's practices on emergency motions, the risk of revealing the plaintiff's best evidence early, and an assessment of how the engineer will present as a witness.
The same pattern runs through the life of a case. Proportionality under Rule 26(b)(1) calls for weighing, among other factors, the importance of the issues, the amount in controversy, the parties' relative access to information, and whether the burden of proposed discovery outweighs its likely benefit. Software can estimate document volumes and review costs, but deciding what to fight for at the Rule 26(f) conference is advocacy informed by strategy. Choosing which claims to plead, which defenses to abandon, and which experts to retain are decisions of the same kind, drawing on experience with how courts and opposing counsel actually behave and on knowledge of the client's tolerance for cost and risk. Clients have always valued these judgments most and paid for them least directly, because they were bundled into hours spent on other work.
Advocacy and negotiation depend on credibility that software cannot supply
Persuasion in a courtroom or a conference room rests on a relationship between people. A judge resolving a discovery dispute is assessing the reliability of the lawyer making representations as well as the merits of the motion, and jurors form views about a case partly through the credibility of the advocates presenting it. Cross-examination is the clearest illustration: questions can be prepared with assistance, but the decision to abandon a line of inquiry after an unexpected answer is made in seconds, in the room, by the lawyer. Tools can help build an argument, test it against counterarguments, and anticipate questions from the bench. They do not stand up and make it.
Negotiation and client counseling work the same way. At a mediation in Meridian, the decisive information may be what Halvorsen's general counsel signals about the company's appetite for a public trial, which no produced document will record. ABA Model Rule 2.1 requires lawyers to exercise independent professional judgment and render candid advice, and it permits them to refer to economic, social, and other considerations relevant to the client's situation. Advising a client to accept a settlement below the theoretical value of its claim, because the litigation is consuming management attention the business cannot spare, is the kind of advice that rule contemplates. Giving it well requires understanding the client as a business and having earned enough trust that the advice is taken.
Firms must rebuild the path by which lawyers learn judgment
The most serious difficulty this shift creates is one of training. Litigation judgment has traditionally been learned through apprenticeship in routine work, as junior associates who spent months reviewing documents absorbed which facts mattered and watched how senior lawyers decided what to do with them. If that work is automated, the informal learning that came with it shrinks. Firms that do not replace it may find within a few years that their mid-level lawyers have supervised AI output without ever developing the instincts needed to evaluate it. Supervisory duties under ABA Model Rules 5.1 and 5.3 make this a matter of professional risk as well as professional development.
Some responses are already available, although none has been proven at scale. Firms can give junior lawyers earlier and more structured responsibility for depositions, hearings, and client calls, with a senior lawyer observing and offering specific feedback afterward. Review of AI-generated work can itself become a teaching exercise if associates are required to explain why a summary or draft is right or wrong, instead of simply correcting it. Leaders might also reconsider how junior lawyers are evaluated, since hours spent on review are a poor proxy for the development that now matters. Each of these measures costs money the billable model does not obviously recover, which is why the question belongs to firm management and not only to practice group heads.
Verification is the foundation for the capabilities that will distinguish litigators
For individual lawyers, the practical question is what to build deliberately. Many of the skills that matter most were once absorbed incidentally through routine work, and as that work is automated they will have to be pursued on purpose. The following capabilities seem most likely to retain or increase their value as automation spreads, because each corresponds to a part of litigation that the preceding sections identify as resistant to delegation.
The discipline to verify machine-generated work efficiently, including checking every authority against its source and noticing when a summary has dropped a qualification that matters.
Working fluency in the tools a lawyer actually uses, sufficient to know which tasks they handle well, where they fail, and what client information may be entered into them consistent with ABA Model Rule 1.6.
The ability to build a case theory from incomplete facts and revise it as discovery proceeds.
Skill in oral advocacy and witness examination, developed through real appearances and not only through preparation.
Enough understanding of a client's business to frame litigation choices in terms of its cost, risk, and commercial objectives.
Little on this list is new, since good litigators have always possessed these capabilities. What has changed is their share of a lawyer's value, and the order in which they must be acquired. The first two items are the ones most lawyers have developed least, and they are a precondition for the rest: a lawyer who cannot verify output cannot safely rely on the tools that free time for strategy and advocacy.
The harder implication falls on how firms charge for their work. If the value of litigation counsel is concentrating in judgment, the billable hour increasingly measures the wrong thing, and firms will face pressure to price matters by phase, outcome, or fixed fee in ways many have resisted [7]. A sensible first step for firm leaders is to examine a handful of recent matters and separate the hours spent processing information from the hours spent deciding what to do with it. That exercise will show how exposed the firm's revenue is to automation. It will also show how much of the firm's real value has gone unpriced. Whether clients will pay for judgment once it is no longer bundled into hours remains the open question for the next several years.
References
Da Silva Moore v. Publicis Groupe, No. 11 Civ. 1279, ECF No. 96 (S.D.N.Y. Feb. 24, 2012). Secondary sources give the reporter cite as 287 F.R.D. 182, but it is not on the fetched page, so verify it before use. https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1:2011cv01279/375665/96
Zach Warren, The Great AI Disconnect: Firms and Legal Departments Are Not Communicating About AI Usage, Thomson Reuters Inst. (Mar. 18, 2026). https://www.thomsonreuters.com/en-us/posts/technology/great-ai-disconnect/
ABA Comm. on Ethics & Pro. Resp., Formal Op. 512 (2024). The article gives the date as July 29, 2024 and discusses Model Rule 1.6. The ABA's own PDF could not be retrieved, so cite the opinion itself after checking it. https://thebarexaminer.ncbex.org/article/fall-2024/generative-artificial-intelligence-tools/
Mata v. Avianca, Inc., No. 22-cv-1461, ECF No. 54 (S.D.N.Y. June 22, 2023). Secondary sources give the reporter cite as 678 F. Supp. 3d 443, but it is not on the fetched page, so verify it before use. This supports only one jurisdiction. https://www.nhd.uscourts.gov/sites/default/files/pdf/Mata-v-Avianca-sanctions-order.PDF
Standing Order Requiring Certification re: Use of Artificial Intelligence (AI) in Filings (D. Colo. Oct. 21, 2024) (magistrate judge chambers order). This is one order only. http://www.cod.uscourts.gov/Portals/0/Documents/Judges/SP/SP_Standing_Order_AI_Filings.pdf
Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7 (2008). The fetched page has no page numbers, so add the pin cite from U.S. Reports. https://supreme.justia.com/cases/federal/us/555/7/
Press Release, Ass'n of Corp. Couns., New ACC Report Finds Generative AI Use in Corporate Law Departments More Than Doubled in a Single Year (Oct. 14, 2025). It is better to cite the underlying report, Generative AI's Growing Strategic Value for Corporate Law Departments. https://www.acc.com/about/newsroom/news/acc-genai-report-corporate-law-departments-ai-use-everlaw
