Anyone watching the legal AI market saw this coming. Kirkland & Ellis, the highest-grossing law firm in the world, is reportedly setting aside $500 million to build its own AI platform rather than relying only on the off-the-shelf tools its competitors can also buy. The number is staggering. The strategy is the part that should change how every firm thinks, even the ones that will never spend a fraction of that.
What Kirkland actually decided
Per reporting on the plan, Kirkland expects to spend more than $100 million this year and hundreds of millions more over the next three to four years developing custom AI services. Chair Jon Ballis framed the goal as taking "the collective intelligence of our institution" and deploying it across the firm. The platform is being built on input from roughly 250 of Kirkland's own lawyers, including about 100 partners, who walked through how they actually work. Outside technology vendors are helping build it, but they will not be allowed to resell it.
Read that last detail twice. The most valuable input is not the model. It is the lawyers describing their real workflows, and the firm refusing to let that advantage become a product its rivals can license.
Why a firm with unlimited options won't just buy Harvey or Legora
The generic legal AI platforms set themselves up as a single answer for every firm. That is exactly the problem. A tool that drops into your word processor can help anyone draft a little faster, but it does not know how a specific deal team runs diligence, where a litigation group's review actually stalls, or which step in a closing checklist quietly burns associate hours every week.
Optimizing a workflow requires being in the weeds of that workflow. You cannot do that from the outside with software built for the average firm, because the average firm does not exist. Kirkland's decision is a clear statement from the top of the market: at some level of seriousness, the way to get real leverage from AI is to build it around how your people work, not to rent a horizontal platform and hope it fits.
Every team saves its fifteen minutes differently
The interesting move in Kirkland's plan is that they are letting their lawyers surface what they want solved, rather than handing everyone the same chatbot and declaring victory. That reflects something every firm owner already knows in their gut: every team has its own way of working, and every team has its own unique way to save fifteen minutes.
Those fifteen-minute savings do not come from a universal tool. They come from automating the specific, repeatable steps that are slow in one particular group: this team's intake form, that team's exhibit assembly, the other team's conflict check. Generic platforms average those differences away. Bespoke workflows lean into them.
You don't need $500 million to apply the same logic
Here is the part that matters for the firms that are not Kirkland. The lesson is not "go build a half-billion-dollar platform." It is that the most sophisticated buyer in the market looked at the build-versus-buy question and chose to build around its own workflows. The logic scales all the way down.
A twelve-attorney firm or a twenty-person accounting practice cannot fund an internal engineering org, and it does not need to. It needs one or two workflows built around how its people actually work, delivered for a fixed price in a few weeks, and owned outright rather than rented back forever. The principle Kirkland is paying $500 million to act on, that custom beats one-size-fits-all, is available to small firms too. The difference is scope, not strategy.
- The most valuable AI input is your lawyers' real workflows, not the underlying model.
- Horizontal platforms optimize for the average firm, which is why adoption fades once the novelty wears off.
- Target the specific fifteen-minute savings inside each team, not a single tool for everyone.
- You can apply Kirkland's build-over-buy logic at any size; the difference is budget and scope, not the underlying idea.
