Kirkland & Ellis committed $500M to building custom AI with Palantir
More than $100 million goes in the first year, aimed at private equity fund formation and compliance documentation. It is the largest disclosed AI commitment by a law firm.
Kirkland & Ellis has committed $500 million to developing custom AI systems with Palantir, with more than $100 million to be spent in the first year, according to the Financial Times. The initial targets are private equity fund formation and compliance documentation.
It is the largest disclosed AI commitment by a law firm, and the choice of workflows tells you the firm has thought about this more carefully than the headline number suggests.
Why fund formation first
Fund formation is the ideal first target for legal automation, and it is not because it is simple.
It is high volume, highly repetitive, and enormously valuable per document. A private equity fund's constitutional documents run to hundreds of pages assembled largely from precedent, where the negotiated variation is concentrated in a limited set of economic terms. The work is expensive because it is voluminous and must be exactly right, not because each instance requires original thought.
That is precisely the shape of task where a well-grounded model earns its keep. It is also work Kirkland does more of than almost anyone, which means the firm has the proprietary corpus to build against — and that corpus, not the model, is the asset.
Compliance documentation has the same properties for the same reasons.
The business model problem it walks into
A law firm that automates document assembly is attacking its own revenue. The billable hour prices input, and a technology that reduces the hours required to produce a fund document reduces what the firm can charge for producing it.
There are only two ways this pays. Either Kirkland converts the work to fixed fees and keeps the margin — which requires clients to accept a price that no longer reflects hours — or it uses the capacity to take on more matters with the same partners.
The second is the more likely intent, and it explains the size of the commitment. Half a billion dollars is not a productivity investment. It is an attempt to change the constraint on how much work the firm can accept, and the firms that get there first take share from those that do not.
Why Palantir rather than a legal vendor
The choice is instructive. Harvey, Legora and a dozen others sell AI to law firms. Kirkland went to a company whose business is integrating models with an organisation's own data under tight access control.
That suggests the firm concluded the hard part is not legal reasoning. It is the data plumbing: getting a model to work against decades of matter files, precedent and client material, with confidentiality walls that hold, an audit trail that survives scrutiny, and the ability to prove which client's information touched which system.
Legal privilege and conflicts of interest are the binding constraints, and they are infrastructure problems. Palantir sells infrastructure.
What is not addressed
Neither firm has said how privilege is preserved when a model trained or grounded on one client's documents serves a matter for another, how the ethical walls are technically enforced, or whether clients have been asked to consent to their material being used this way.
Those questions have answers, and law firms have handled analogous issues with knowledge management systems for years. But the disclosed commitment is $500 million and the disclosed governance is nothing, and it is the second of those that determines whether this survives its first conflicts challenge.
Runs the newsroom. Rename this profile in the studio to your own byline.
Related
Every weekday, the AI stories that moved money or shipped code.
No cross-posting, unsubscribe anytime. See all newsletters