For the past two or three years, discussions about artificial intelligence (AI) have been taking place in the glass-and-steel hallways of large law firms in Chicago, New York, and London. These discussions have alternated between sincere enthusiasm and subdued concern about the technology’s potential effects on a field that depends on billable hours and human judgment.
With $500 million of its own revenue dedicated to developing a unique AI platform, beginning with $100 million in 2026, Kirkland & Ellis, which recorded $10.6 billion in revenue last year and has never shied away from operating at scale, has now taken the discussion to a more tangible level. The news is the biggest AI investment pledge made by a legal firm, and it should be taken seriously for what it represents as well as what it means for the future of the sector.
The decision to build instead of purchasing is the one that should be carefully considered. Law firm technology executives were usually opposed to bespoke development five years ago, according to Andrew Johnson, chief information officer at Brownstein Hyatt, because the expense, complexity, and uncertainty all pointed toward licensing from reputable suppliers. The computation has changed. More than 180 technological experts from both inside and outside the firm will contribute to the development of Kirkland’s platform, which will be sculpted by 250 of the company’s own attorneys.
This suggests a really customized solution rather than a rebranded off-the-shelf tool. Given the competitive stakes involved, the company’s decision to withhold the name of the underlying generative AI model the platform will rely on is a particularly cautious one. Alongside its unique build, it will continue to license some third-party programs, suggesting a hybrid strategy rather than complete vertical independence.
It is easy to understand the strategic reasoning. General-purpose legal AI technologies are unlikely to meet the specialized demands of a company with $10.6 billion in sales and practices that include private equity, M&A, restructuring, and litigation. The nature and volume of the document review processes, contract analysis patterns, and research questions that emerge in a Kirkland matter differ from those of a mid-sized regional business.
Developing something that serves as a true competitive edge instead than table stakes technology that every company in the market will have access to on the same conditions requires building to those specs using proprietary training data derived from years of actual Kirkland work output. Freshfields has adopted a similar strategy, collaborating closely with Anthropic to create legal applications; this is a partnership model as opposed to a completely custom construct, but it is motivated by the same basic principles.
For the past two years, the dangers have been demonstrated in courtrooms. In scores of cases, US federal judges have sanctioned attorneys for submitting AI-generated research with fake citations, including quotes that were never written, cases that were never decided, and legal authority that don’t exist. In the legal field, the consequences of a hallucinated source include not just humiliation but also professional reprimand, harm to clients, and in certain situations, severe penalties.
Before it even approaches a filing or a negotiation, any custom AI platform designed for litigation or transactional work must address this type of error. This is probably why Kirkland is investing the kind of money that it is taking three to four years to build rather than shipping something quickly.
It’s also unclear how other big businesses will compete or whether the investment will increase productivity and profit margins enough to warrant the size of the commitment. As you watch the announcement, you get the impression that Kirkland is purposefully stating that it wants to control the technology path of its own practice instead of waiting for the market to provide it. It will take years to see whether the platform performs as effectively as the investment suggests. However, there is no question about how serious the wager is.
