The world’s highest‑grossing law firm is shunning off‑the‑shelf tools to capture its “collective intelligence” – and equity partners are funding the bet out of their own pockets
Tom Borman, LawFuel contributing editor
The legal tech arms race just moved up a weight class. Kirkland & Ellis, the world’s highest‑grossing law firm, has set aside an eye‑watering $500 million to build its own proprietary generative AI platform, rather than relying on the same off‑the‑shelf tools everyone else can buy.
The strategy, revealed by firm chair Jon Ballis and first reported by the Financial Times, marks a deliberate pivot away from simply licensing commercial software. Ballis says the firm expects to spend more than $100 million this year alone on custom AI services, with hundreds of millions more to follow over the next three to four years – roughly 1% of Kirkland’s annual revenue.
“We don’t get hired for the floor”
While the rest of BigLaw scrambles to ink licensing deals with vendors such as Harvey and to strike alliances with the likes of Anthropic and Google, Kirkland has come to a blunt conclusion: if everyone can buy the same AI, it stops being an advantage.
Ballis has been explicit that widely available AI tools are raising baseline standards across the industry – but that is not the level on which Kirkland competes.
“The idea is that we’re going to take the collective intelligence of our institution and be able to deploy that throughout our firm… [Widely available AI tools are] raising the floor for everyone, but we don’t get hired for the floor.”
In other words: if generative AI becomes table stakes, Kirkland intends to own the table.
Inside the $500 million build
Kirkland is not coding this thing in a vacuum. The firm is throwing bodies and budget at the project – and locking down the intellectual property.
- The development muscle. The platform is being developed with input from around 250 lawyers, including 100 partners, alongside 180 technology professionals, who are mapping real‑world transactional and disputes workflows into the system.
- IP lockdown. External technology companies are helping to build the architecture, but they are contractually barred from reselling the platform or its components to rival firms. Kirkland will retain full ownership of the finished product.
- End‑to‑end ambition. Rather than forcing lawyers through a patchwork of point solutions for document review, due diligence and drafting, the goal is a platform that can support multi‑step, complex legal mandates from initial scoping through execution.
The vision is not just a smarter search box, but a firm‑wide operating system tuned to Kirkland’s own playbook.
Financial firepower – and the partner bill
The numbers behind this are as much the story as the technology. Kirkland has recently been reported as the first law firm to smash through the $10 billion revenue barrier, hitting about $10.6 billion in global revenue, according to multiple reports citing the firm’s latest financials. Its profits per equity partner have been reported at a record $11.1 million, up roughly 20% year‑on‑year.
That kind of balance sheet makes a $500 million AI bet possible – but not painless. Ballis has indicated the spend will be funded out of current revenues, implying a short‑term dent in partner distributions as the platform is built out.
For equity partners used to eight‑figure profit numbers, this is the cost of trying to future‑proof the franchise.
Two emerging AI camps in BigLaw
Kirkland’s move sharpens a fault line that has been forming inside the upper reaches of BigLaw: the divide between firms that buy AI and firms that build it.
| Strategy group | Notable firms (illustrative) | Core approach | Key advantage / risk |
|---|---|---|---|
| Buyers & partners | Magic Circle, Wall Street elites (e.g. Freshfields, other global firms) | Large‑scale licensing of platforms like Harvey, along with deep alliances with Big Tech and frontier AI labs. | Rapid rollout and lower capex, but tools – or close equivalents – can eventually be sold to rivals. |
| Bespoke builders | Kirkland & Ellis; other firms experimenting with in‑house tools | Heavy upfront investment in proprietary platforms trained on internal data and workflows, often on dedicated or tightly controlled cloud infrastructure. | Potentially uncopyable data advantage and tighter control, but significant execution and R&D risk. |
Freshfields’ recent co‑development partnership with Anthropic – under which Claude will be deployed to thousands of staff and legal AI tools may later be sold to other firms – is emblematic of the “buyer/partner” camp. Kirkland, by contrast, is building for itself and itself alone.
Does this kill the billable hour – or just wound it?
The most interesting part of Kirkland’s AI push may not be the tech stack, but the pricing model it is designed to support. Ballis has already flagged that the platform will accelerate a shift away from the traditional billable hour towards value‑based pricing, with matters priced on outcomes and commercial value rather than hours spent.
If the platform delivers on its promise – compressing timelines and stripping out low‑value associate work – Kirkland will be under even more pressure to move clients off pure time‑based billing. That is a feature, not a bug: the firm plainly expects to monetise speed, not slowness.
This, of course, sits against a backdrop of ongoing “AI hallucination” horror stories in UK and US courts, where poorly supervised tools have generated fictitious case law and sanctions. Kirkland’s answer is to spend its way towards total control over the data, the workflows and the risk envelope, selling clients on a story that its AI is not only faster, but safer and more accountable.
For the rest of the Am Law 100 and Magic Circle, the uncomfortable question now writes itself: in an AI market where anyone can sign a licence, what are you building that your competitors can’t simply buy?
