The legal profession has spent the past three years talking about artificial intelligence the way Victorian aristocrats discussed electricity: with fascination, mild alarm, and absolutely no intention of rewiring the house themselves. A&O Shearman has now called time on that comfortable ambivalence.
The transatlantic mega-firm, born from last year's merger of Allen & Overy and Shearman & Sterling, announced Thursday that it will invest $1.1 billion over five years to embed AI across its global operations. More striking than the figure is the mechanism: every practice group—from M&A to structured finance to litigation—must commit at least 3% of billable time to developing AI tools. That's not a suggestion from innovation consultants. It's a mandate from management.
The economics of self-disruption
The math is deliberately uncomfortable. A&O Shearman billed approximately $4 billion last year. Three percent of billable hours across the partnership represents roughly $120 million annually in foregone revenue—time partners and associates will spend building software instead of billing clients. The firm is betting that the productivity gains and competitive moat will more than compensate.
The investment will fund dedicated AI engineering teams in London, New York, and Singapore, partnerships with foundation-model providers (the firm already uses a customized version of Anthropic's Claude for contract analysis), and what the firm calls "practice-specific co-pilots"—bespoke tools trained on the accumulated expertise of individual practice groups. The capital commitment dwarfs anything previously seen in legal technology; the prior record was Baker McKenzie's $400 million digital transformation program announced in 2024.
Why now, and why this firm
A&O Shearman's position is unusual among elite firms. The merger created a genuine transatlantic footprint with deep pockets, but it also created integration headaches and partner anxiety about cultural dilution. A massive AI bet serves multiple purposes: it gives the combined partnership a shared mission, justifies the merger's scale thesis (you need $4 billion in revenue to afford $1.1 billion in R&D), and positions the firm as the obvious choice for clients who want their outside counsel to actually understand the technology reshaping every industry.
The timing is also pointed. Tuesday's midterm results suggest a regulatory environment more skeptical of unchecked AI deployment—Democratic majorities in both chambers have already signaled interest in comprehensive AI legislation. A&O Shearman is betting that the coming wave of AI compliance work will be enormous, and that firms who built the tools will win the advisory mandates.
The partnership question
The 3% mandate is where this gets interesting. Law firm partnerships are famously resistant to anything that reduces individual partner autonomy or short-term compensation. Telling a capital markets partner that she must spend 3% of her time—time that could otherwise generate $2,000-an-hour fees—on building software requires either genuine conviction or genuine coercion.
A&O Shearman appears to be deploying both. The firm has tied a portion of partner compensation to "AI engagement metrics" and created a new category of "technology partner" with equity stakes for lawyers who lead successful tool development. It's a recognition that the billable hour, the profession's organizing principle for a century, may not survive the decade.
Our take
This is either the moment Big Law finally grew up about technology or an extraordinarily expensive way to discover that lawyers are not software engineers. The smart money is on the former. A&O Shearman's leadership has correctly identified that the moat around elite legal services—expertise, judgment, institutional knowledge—is precisely what large language models are learning to replicate. The only sustainable response is to own the tools rather than be displaced by them. The rest of the Magic Circle has perhaps eighteen months to respond before the gap becomes unbridgeable.
