The Cost of Faster Legal Work

Welcome to Attorney Intelligence, the weekly newsletter from PointOne where we break down the forces reshaping legal from the inside out.

Last week, I argued that a growing share of legal work has become predictable enough to be priced like a product. Taking a step back, I want to break down the cost of delivering that product and how AI is impacting this cost.

Salaries, occupancy, and technology are the three major areas of law firm spending. By this assessment, technology is the fastest-growing of those categories, particularly since the pandemic. I’m focusing on compensation and technology today because AI changes the relationship between them: firms are investing in tools that can reduce production time while continuing to pay the lawyers who perform and supervise the work.

To understand the return on that investment, we first need to understand the costs already in place.

What Does a BigLaw Associate Actually Cost?

An associate’s salary is only part of the firm’s investment. The firm also pays to recruit, train, support, and supervise that lawyer, then needs to collect enough revenue from their work to cover those costs.

Firms calculate this in different ways. Some allocate overhead to each lawyer. Others focus on the profitability of a matter, client, or practice group. The basic inputs are consistent: compensation, benefits, operating expenses, billable hours, rates, discounts, and collections. Direct expenses of employing labor consumed 32% of the average firm’s revenue in 2025.

What a first-year associate salary tells us

In 2025, the median first-year salary was $215,000 at firms with more than 700 lawyers, and six major markets had reached a median of $225,000. Large U.S. firms typically benchmark associate pay against the Cravath scale, which sets expectations for compensation by class year. Competitors often respond to salary increases because they recruit from the same candidate pool.

Once a lawyer is hired, that salary is largely committed over the budget period, even when the time required for individual assignments changes. The cost attributed to a matter therefore depends on how the lawyer’s time is allocated and how much of that work the firm can bill and collect.

What the firm pays beyond salary

Beyond salary and bonuses, firms pay employer payroll taxes, benefits, and the operating expenses that support legal work. Recruiting and development also add summer programs, search fees, onboarding, professional training, and senior-lawyer supervision that are difficult to break down on an industry-scale. 

In 2026, the federal payroll-tax rules stated that an employer must pay 6.2% Social Security tax on wages up to $184,500 and 1.45% Medicare tax on covered wages. On a $225,000 salary, those taxes total approximately $14,702 before unemployment insurance or workers' compensation.

Due diligence and repeatable transaction work as early targets for AI, while experienced lawyers remain essential to reviewing the results. Supervision has arguably one of the greatest opportunity costs that law firms must factor into their  associate expenses. Partners spend time reviewing and correcting work that they could otherwise devote to client matters or business development. 

An established approach, illustrated in Altman Weil’s profitability model, combines compensation and allocated expenses with expected billable hours and collections to assess the rates needed to cover costs and support a target margin. However, one assumption the model makes is that partners and associates use similar resources, including a similar technology package, making those costs relatively straightforward to allocate across lawyers.

Technology as a previously predictable cost to calculate profit

Previously, much of the technology budget supported that assumption. Firms budgeted for document management, legal research, timekeeping, billing, and security systems through subscriptions priced by firm or user. 

Costs still rose with new hires, new software, and contract renewals, and some services carried usage charges. But much of the core technology budget could be planned around the number of lawyers and systems the firm needed to support.

AI adds to that existing commitment through both adoption costs and, increasingly, charges tied to use. Consumption pricing is a marked change in that calculation. Legora’s consumption-based pricing for Agent Pro, for example, allows usage to be tracked by organization, user, and project. Firms therefore need to account for variable use alongside their fixed commitments.

What AI adds to the cost equation

AI changes this cost structure in two ways. It can reduce the hours required for research, summarization, clause extraction, diligence, and routine drafting. The potential time savings are substantial, with expectations of freeing nearly 240 hours annually.

AI also introduces new expenses for licenses, integration, data preparation, workflow design, training, security review, and human verification. Some of these costs occur during implementation, while others recur or increase with use. Significant specialist support often accompanies the tools. Firms may need data architects, data scientists, prompt engineers, and legal technology specialists, alongside additional cybersecurity and insurance spending. Client information also needs to be kept separated and protected.

“The cost of AI goes beyond the software. Firms need people to prepare their data, build useful workflows, and protect client information.”

— Director of Revenue and Financial Systems at AmLaw 200

The initial adoption process resembles an R&D investment. Lawyers and technical staff must test tools, prepare knowledge sources, design workflows, and evaluate results before the firm can determine whether AI improves profitability. That work consumes time that could otherwise support client matters or other operational priorities.

Ropes & Gray’s TrAIlblazers program provides a clear example. The program allows first-year associates to apply part of their billable-hour expectations toward AI training, simulations, and experimentation. The firm is investing associate capacity in developing new skills before expecting those skills to improve client delivery.

A useful AI budget separates initial development and implementation from recurring licenses, variable usage charges, maintenance, training, and review. Firms can then evaluate the cost of the complete workflow, including the time required to verify and correct AI output, against any improvement in cost, quality, or turnaround time.

The biggest challenge for law firms is that they’re currently carrying both cost bases. In 2025, average law-firm technology spending rose 9.7%, knowledge-management spending rose 10.5%, and lawyer-compensation spending rose 8.2%, while lawyer headcount increased 2.9%, according to the 2026 Report on the State of the US Legal Market. These numbers demonstrate that AI has not yet produced a simple substitution of software for lawyers.

The cost of associate turnover

The NALP Foundation reported an average associate attrition rate of 19% in 2025 among participating U.S. and Canadian firms, with 83% of departures occurring within five years of hire. Associates often leave by their third year, requiring firms to repeat the investment in recruiting and training replacements.

That timing can limit the firm’s return. In archived Canadian Bar Association guidance, new lawyers begin contributing to net profit around three years in. An associate becomes profitable under a fully allocated model when revenue attributed to their work exceeds their compensation, development costs, and assigned share of firm expenses over the period measured. As associates become more efficient and require less supervision, fewer hours may be written down by partners before billing, allowing more of their work to translate into collected revenue. Early departures can therefore limit the firm’s return just as that investment begins to pay off.

Overall attrition declined slightly from 20% in 2024 to 19% in 2025. NALP’s public findings do not attribute that decline to AI, although its latest study examines whether access to AI support and training influences associates’ decisions to leave. This introduces another potential return on technology investment: could AI reduce the burden of routine work and improve associates’ experience enough to keep them at the firm longer?

What firms should measure

Time tracking remains essential even when billing moves away from hourly rates. Firms still need to know how much lawyer time a matter requires to assess its margin. Some firms are adding fields to time-entry systems to identify AI-assisted work, giving them a starting point for comparing how similar assignments are completed.

“Time is a law firm’s raw material. Even if AI changes how we price legal work, we still need to measure what it takes to deliver it.”

— Director of Revenue and Financial Systems at AmLaw 200

A firm needs a consistent internal model that connects the cost of employing its lawyers to the work they complete and the revenue it collects. It should distinguish annual cash commitments, shared expenses allocated to a lawyer or matter, and costs that increase when more work is performed. 

The next step is to follow the saved time. Was it reassigned to paying work, invested in training or business development, or pro bono work? Did the firm collect more revenue with its existing team, avoid a hire, reduce replacement costs, or improve the return on a fixed-fee matter? Nonbillable uses may have value, but their return should be assessed over an appropriate period.

There’s a major revenue opportunity in freed capacity. Understanding the use of saved time is central to evaluating AI’s return: the firm needs to understand whether faster delivery allows its existing team to serve more client demand.

Taken together, changes in time, staffing, technology spending, and collections give leaders a basis for deciding which workflows to expand and how to staff and price them. They also help firms explain the benefits they can share with clients and the costs they still incur to deliver the work.

Legal Bytes

Harvey targets lower AI delivery costs. Harvey introduced Tenet, its first post-trained open-weight model, and reported a 90% reduction in cost per query and three times more intelligence per token. Harvey AI.

Wall Street wants its share. Goldman Sachs, Morgan Stanley, and Citigroup are pressing major firms to explain how AI savings affect fees. Modern Counsel.

AI native firms test fixed fee economics. New legal providers are using AI-centered delivery and fixed fees to test whether lower production time can support a different margin model. Artificial Lawyer.

Thanks for reading,
Katon