A head of innovation at one of the largest US firms said something to me this year that I haven't been able to shake: "As we pay for AI that makes us faster, we're paying for our own revenue to go down."
Most firm leaders understand this privately. Very few say it in a partnership meeting, because it sounds like an argument against AI, and nobody wants to be the partner arguing against AI in 2026. The concern is economic, and the firms that work through it deliberately will come out of the next few years in a much stronger position than the ones that let it happen to them.
The arithmetic partners already see
The structure of the problem is simple. The firm pays for AI, often per seat or by usage. Lawyers use it, and routine work takes fewer hours. If the matter is billed by the hour, fewer hours means less revenue on that matter, unless the rate changes or the firm prices the work differently. The firm has paid to reduce its own top line.
In partnership meetings this rarely gets framed as a pricing problem. It shows up as a question about utilization, a practice group billing fewer hours on routine work, or a partner asking why the AI line item keeps growing. Each of those is the same issue seen from a different chair. A leader who names it directly gives the partnership something it can actually decide.
This is not a small corner of the business. The Thomson Reuters Institute's 2026 Report on the State of the US Legal Market found that "90% of legal dollars still flow through hourly billing arrangements," even as firms invest heavily in AI. Almost every dollar of AI efficiency lands on a pricing model that was built to reward effort.
The same head of innovation put the second half of the problem this way: "We cannot have a competitive advantage by paying retail like everybody else." If every large firm buys the same tools on the same terms, the efficiency is shared across the market and eventually passed to clients. The firm pays for the tool and keeps very little of the gain.
What firms are doing about it
From what I see inside firms, the leaders who are ahead on this are doing some combination of four things.
They're pricing the work where AI changes the effort. Repeatable work, such as first-pass diligence, standard agreements, and regulatory filings, is moving to fixed or capped fees. The firm keeps the margin when AI reduces the effort, and the client gets predictability. This works best when the firm has real data on how long the work takes today, which is why measurement comes first.
They're treating efficiency as capacity. A practice group that does the same work with fewer hours can take on more matters with the same team. That only helps revenue if the demand is there and the partners go get it, so this is a business development decision as much as a technology one.
They're watching how the AI itself is priced. Per-seat and per-token pricing grows with adoption. The more lawyers use the tool, the more the firm pays, while hourly revenue on the same work falls. Firms that understand this are asking their CFO to model AI spend against matter revenue, and they're pushing vendors toward fixed costs that don't rise every time a lawyer finds a new use.
They're deciding what the firm keeps. The head of innovation's full point was that the firm's expertise flows into these tools, and the firm then pays to rent that knowledge back. The firms thinking furthest ahead want the knowledge their lawyers produce to accumulate in something the firm owns, so the advantage stays with the firm.
Our AI adoption playbook for firms that bill by the hour covers how to roll AI out inside these constraints. This piece is about the economics underneath the rollout.
The asset is what the firm knows
When the effort to produce legal work falls, the value of the firm shifts toward what it knows: the positions it has taken, the clauses it has negotiated, the judgment of its partners about what holds up and what doesn't. Clients have always paid for that judgment. The hourly bill was the way the firm charged for it.
That means the question for a managing partner is bigger than which tool to buy. It's whether the firm's knowledge is organized so the firm can price it, protect it, and deliver it faster than the hours used to allow. We wrote about the retention side of this in when a partner leaves, the firm should keep what they knew. The same logic applies to every matter the firm closes.
What leaders can do this quarter
- Measure where AI changes the effort. Pick the matter types where lawyers already use AI and compare hours before and after, by task. Without this, every pricing conversation is a guess.
- Choose two matter types for a pricing pilot. Work with the relationship partners on clients who have asked for predictability. A small, well-chosen pilot gives the partnership real numbers to react to.
- Put the AI cost model on the CFO's agenda. Model per-seat and usage-based spend against the revenue it touches, and see what happens as adoption grows.
- Decide where the firm's knowledge accumulates. If the answer is inside vendors' products, that's a strategic choice, and it should be made on purpose.
- Talk to clients before they raise it. Clients know AI changes the work. A firm that brings a pricing proposal to that conversation gets to set its terms.
How we approach it at Reframe
We price Reframe as a fixed annual license, never per seat and never per token, so the firm's cost doesn't rise as lawyers use it more. The platform runs in the firm's own cloud, and the model it builds of the firm's matters, clients, and documents belongs to the firm. As lawyers work, what they know accumulates in a system the firm controls, and every answer is cited to the document it came from.
If your partnership is starting to ask these questions, we're glad to share what we're seeing across firms.
Drafted with Reframe Editorial Agent, an AI employee at Reframe. 