The ontology gap, why no law firm has answered Kirkland & Ellis yet

What Kirkland actually built

Most coverage has described Kirkland's partnership with Palantir as an AI platform. That undersells it. What Kirkland is building is an ontology: a structured model of the firm's own world, its funds, investors, side letters, clause positions, and the decisions it has made about each of them, with an agentic layer that reasons over that structure in the flow of real work.

The distinction matters more than the price tag. A point solution like Harvey or Legora is trained largely on public material, court filings, judgments, legislation. It produces competent, generic output, and for routine drafting that's genuinely useful. But the work that defines any specialist practice doesn't live anywhere public. Limited Partnership Agreements aren't filed with courts, and neither are the heads of terms a property team negotiates or the settlement positions a disputes team reaches. The positions that became a firm's own market standard, on fund documents, lease clauses or liability caps, were never scraped or indexed, because the only institutions holding that knowledge are the firms that built it.

How the ontology compounds over time

An ontology captures that knowledge as structure rather than as documents in a folder. Clients, matters, positions taken on recurring points and the outcomes a firm has accepted become connected objects with defined relationships, not files to search through. For Kirkland, those objects are funds, investors, side letters and clause positions. For a corporate team they would be deal terms and precedent positions; for a property team, lease structures and title issues; for a disputes team, case strategy and settlement history. The mechanics hold regardless of practice area. A junior lawyer works inside a system that already knows the client's history, the firm's standard position on each recurring point, and the range of outcomes the firm has accepted before. The lawyer stays the decision-maker. The system removes the latency between the question and the answer.

The ontology also bakes reasoning and decisions into your knowledge structure. A legal document contains the positions that were negotiated and agreed upon by parties. It is like reading the last page of a novel: you know what happens at the end, but you’ve learned nothing about the characters’ development and the plot twists that led to the outcome. Today, if a position deviates from the firm’s recognized standard the rationale is not captured. It lives in the inbox or the head of the lawyer who negotiated it. Off-the-shelf AI tools cannot hope to learn from or replicate this, nor can they capture it. Yet that reasoning is what forms a firm’s institutional knowledge. It is precisely where their value proposition lives.

The ontology captures decisions like an audit trail. It records the full history of the ‘who’, ‘when’ and ‘why’ behind each clause in a document. This knowledge is structured into the infrastructure which AI models are trained on. It accelerates junior lawyers, so they can efficiently reason like a partner. It prevents reinforcing any biases that existed in historical precedent or singular negotiations which are not broadly applicable.

The part that should worry every firm watching from the sidelines is what happens over time, in any practice. Every matter that runs through the ontology enriches it, whatever kind of matter it is, captured as structured data rather than lost in a document store. Five years in, a firm's own system will encode its own positions, client relationships and risk judgement in a form no generic model will ever replicate, because no generic model has access to that corpus.

The choice every firm eventually faces

That's what makes an ontology an asset rather than a tool. A point solution is available to every competitor with a chequebook, and it makes lawyers faster at the same commodity work everyone else is also getting faster at. An ontology gets harder to copy the longer a firm runs it, which is exactly why every firm exposed to competitive pressure on high-value, repeat work, not just those with heavy funds or M&A practices, should be paying closer attention to this deal than the headlines suggest.

We see the same fork opening up in every sector we work in, financial services, insurance, parts of the public sector. The institutions treating their own knowledge as an asset worth structuring, rather than a cost to automate away, are the ones pulling ahead. Law is arriving at this fork later than most industries, but the decision in front of it is the same one every sector eventually has to make.

Taking that leap is a matter of conviction, a partnership prepared to back a multi-year infrastructure decision rather than a twelve-month tool rollout, and willing to treat what the firm already knows as worth building on. None of that requires Kirkland's budget, and no firm has to work out how to start on its own. The firms bold enough to make that call now will spend the next five years compounding an edge that's theirs alone. Everyone else will spend it renting the same tools as the firm next door.