In the first three months of 2026, partner hiring at large U.S. firms reached its highest level in six years. In 2025, the number of partners firms hired away from other firms rose about 18%, and the number of associates hired away rose about 17%.
The entry-level side of the same market ran the other way. In the 2025 recruiting cycle, the median number of summer associate offers that each law office extended fell to four, an all-time low in the records of the National Association for Law Placement (NALP). And firms of more than 500 lawyers hired about 540 fewer entry-level associates from the class of 2025 than from the class before.
As a result, the legal job market is splitting in two. Firms are paying for experienced talent while hiring fewer entry-level lawyers to train. Many firms expect AI to handle the routine tasks that previously justified large first-year associate classes, though current data has not yet confirmed this trend.
The split is happening inside a growing profession. U.S. legal-sector employment hit a record at the end of 2025 and has kept climbing, to about 1.25 million jobs in September 2026 on preliminary Bureau of Labor Statistics (BLS) data.
If you run a firm or a practice group, the numbers put three decisions in front of you. The first is how large your first-year class should be. The second is who pays for junior training if the billable work that funded it goes to software. The third is how you document who checked the first-pass work.
- Large U.S. firms hit a six-year high for partner hiring in early 2026, while entry-level hiring fell 7.5%.
- AI may reduce junior work, but billable hours rose 4.2%, leaving the cause and staying power unsettled.
- Leaders should set 2028 class size and training budgets together, fund supervised learning and document first-pass review.
- Acting now can protect the talent pipeline, control training costs and prepare for demand for lawyers and technologists.
The Entry-Level Contraction
Entry-level hiring at the largest firms fell for the first time in more than a decade. Those 540 fewer hires are a 7.5% drop from the class of 2024 peak. NALP's report named possible hiring shifts driven by generative AI as one of several factors behind the year's declines, alongside the federal hiring freeze and shrinking public-interest funding, but it did not say how much AI contributed.
There is a second possible explanation, and it predates the AI forecasts. In April 2024, NALP described the lateral market as "still in a course correction after the over hiring that took place in 2021," and suggested many firms may still be dealing with excess capacity, particularly at the associate level.
The case that AI is the cause rests on the firms' own forecasts. In Citi and Hildebrandt Consulting's December 2025 outlook, 86% of large firms planned to grow their associate ranks through 2027, yet only 35% planned larger first-year classes and 37% planned more summer associates. Firms in that outlook expect AI to automate the repetitive work that falls to the least experienced lawyers.
By the summer of 2026, Citi's survey of 57 large firms found that two-thirds expect fewer first-year associates in 2028 than in 2025 and nearly 60% expect fewer second-years, while 19% do not yet know how many first-years they will hire in 2028.
The measured results point the other way. Billable hours across the law firm industry rose 4.2%, by Citi's count, in the first half of 2026 against a long-run average of 1.5% to 2%, so AI has not yet reduced the hours clients buy.
The work firms expect the tools to absorb is the routine work that falls to first-years, which at the largest firms often means basic legal research and drafting memos and other documents. Citi's survey also found that associates use AI daily at two-thirds of firms. But none of the reporting behind these numbers names a firm that cut a first-year class because software now does the first pass.
Citi and Hildebrandt describe the shift as one from a pyramid toward a cylinder: a slimmer base and more reliance on senior lawyers.
The Training Pipeline Challenge
Historically, law firms trained senior lawyers as a side effect of billing clients for junior associate hours. For instance, a first-year associate might spend a year on document review. The client paid for that time, and the firm gained a second-year associate with practical experience.
Clients paid for junior training because junior hours were billed. If software takes those hours, who pays depends on how your firm bills. About half of the 57 firms in Citi's survey say AI has already affected the billable hour and three-fourths expect it to by 2028.
ABA Model Rule 5.1, which each state adopts in its own version, requires partners to make "reasonable efforts" to ensure the firm maintains measures so all lawyers adhere to ethical rules. Direct supervisors share this same duty. ABA Formal Opinion 512, issued in July 2024, applies that duty to generative AI: it does not change when first-draft or review work moves from a junior associate to software.
Four practical steps follow from this shift. First, set the 2028 entry-level class size and the training budget together, with a number for each.
Second, directly allocate funds for training that was previously billed to clients. Offer billable-hour credit for supervised introductory work performed for educational purposes rather than client billing, as well as for the time senior lawyers spend reviewing and guiding that work.
Third, maintain detailed records per matter identifying whether a team member or software completed the initial work, along with who reviewed it. That record shows reasonable supervision if a client or regulator asks, and shows which juniors are getting real training work.
Fourth, ask lateral candidates about their training background and the specific nature of their past work experience. A firm that recruits experienced lateral talent instead of developing entry-level attorneys relies on other firms to have trained them.
What Clifford Chance and Baker Botts Show
In October 2025, Clifford Chance told about 550 business-services staff in its London office, in finance, HR and technology, that it expected to cut about 50 roles and change about 35 more. That is roughly 10% of those roles in that office. Staff were told that greater use of AI and weaker demand for some support work were behind the plan, along with work moving to lower-cost offices.
So far, the law firm cuts attributed to AI have hit support staff, not lawyers.
Baker Botts is betting the other way. Its managing partner, Danny David, says the firm will hire more associates. He argues that when AI cuts the hours a matter takes, a profit model built on associate volume is more exposed than one built on specialized expertise. He also says AI has already improved associates' prospects because new hires grew up with the tools. He adds that some firms may see layoffs while his will not.
By its own reporting, Baker Botts hired 84 entry-level associates this year, a minor decline from 86 the previous year. Because its strategy centers on future incoming classes, current figures do not fully resolve the debate.
Rising billable hours support the bet, because firms with more work need more associates to do it. Falling entry-level hiring supports the Citi forecasts of smaller first-year classes in 2028. The evidence so far is split.
Who Is Actually in Demand
BLS put paralegal unemployment at 2.0% on average in 2025, against 4.3% for all workers and 0.8% for lawyers. A low unemployment rate does not mean a growing occupation. BLS expects demand for paralegals to be limited by advances in AI because those tools are expected to make paralegals more efficient at research and document preparation, which may reduce demand. It projects about 40,800 openings a year, almost all from people leaving the occupation.
Demand is strongest in legal technology. Large firms are hiring AI directors and legal-technology staff at $200,000 to more than $400,000. In July 2026, LinkedIn showed at least 16 top firms looking to fill more than 25 such jobs in total (among them Latham & Watkins; Gibson, Dunn & Crutcher; and Pillsbury Winthrop Shaw Pittman). Covington & Burling was offering up to $438,000 for a director of applied AI.
Kirkland & Ellis says its AI team has more than 180 engineers and data scientists, and in June 2026 its careers board listed nearly 40 open AI and legal-tech jobs.
These hires are directors and engineers rather than paralegals. The lawyer growth Citi and Hildebrandt expect is at the senior and mid levels. So firms are adding technologists and senior lawyers.
When recruiting paralegals today, prioritize candidates capable of auditing and verifying AI-generated output. Once software drafts, checking its work is the job.
The Open Question for the Profession
In Citi and Hildebrandt's December 2025 outlook, 63% of large firms expected AI to bring some change by 2035 to their leverage model, the ratio of salaried lawyers to the partners who own the firm.
If past declines are a guide, this one is not over. In NALP's data going back to the 1980s, big-firm entry-level hiring has hit bottom about three years after each peak, and each year's decline on the way down was larger than the one before it. The case for a milder decline this time is demand, since firms are billing more hours than usual.
A second reason has nothing to do with AI. The classes graduating in 2025 and 2026 are smaller than the class of 2024, so firms have fewer graduates to hire from.
If AI takes the hours that trained the partners of 2026, firms have to decide now who trains the partners of 2040 and who pays for it.