A new report from the AI Futures Forum finds that AI adoption across UK law firms has stopped paying off, and it places responsibility for the shortfall squarely with firm leadership rather than IT departments or technology vendors.

The report, "The AI Value Gap: The Human Economics of Stalled AI Transition in the Professional Services Sector," published in late September 2026, follows the forum's earlier research tracking AI uptake across the legal sector since mid-2025. Its central finding: usage curves are climbing, budgets are growing and tools are proliferating, yet measurable returns remain elusive for most firms.

91% and Falling Short

The report draws on Thomson Reuters' 2026 Future of Professionals research, which found that 91% of fee earners believe their own organisations are falling short of what AI could realistically deliver. The problem, the report argues, is not under-investment. Firms with a genuinely visible AI strategy are nearly four times more likely to see measurable return than those without one — yet most firms and legal departments still cannot say what their return actually is. Only 18% collect ROI metrics in any systematic way.

The report also highlights a striking anomaly: in several major US firms, billable hours are rising even as AI use grows, and nobody has offered a convincing account of where the promised time savings went.

A Confidence Gap, Not a Technology Gap

James Tuke, CEO of the AI Futures Forum and the report's author, dismisses the instinctive explanations — immature tools and lagging governance — as insufficient to explain a gap of this size. "The value gap is a confidence gap, and it sits squarely with leadership, not IT or HR," he writes.

The report's evidence will make uncomfortable reading for managing partners. Thomson Reuters' Stand-out Lawyers survey found that even among the heaviest partner users of AI, only around a third have discussed its impact with most of their clients. Partners, Tuke argues, are a firm's commercial transmission mechanism — the people who set matter strategy and staffing expectations. A partner who has not personally used the tools enough to talk about them credibly cannot sell that impact to a client, however good the firm's official strategy claims to be.

The Ownership Question

The report identifies a more self-interested reason for partner hesitation. As firms train AI systems on their partners' accumulated judgement, a question of ownership arises: who benefits from the resulting asset? Neither the billable hour nor a profit share pays a partner for supplying the raw material of a system that keeps generating value long after retirement.

Tuke points to Kirkland & Ellis's decision to commit US$500 million to a proprietary platform, which commentators read as being about ownership and control, not efficiency. Some firms are now weighing whether to spin off their AI capability as a separate commercial entity — in effect, becoming their own "NewMod" rather than being disrupted by one.

Foundation Labs Enter the Application Layer

The competitive threat is no longer confined to AI-first start-ups, the report warns. In 2026, Anthropic launched Claude for Legal, with direct connectors into major vendors and deployments already live at Freshfields and Quinn Emanuel. The move signalled that foundation labs themselves are moving into the application layer built on top of them. This month, OpenAI followed suit with the launch of Astra for Law. Firms comfortable with their current vendor relationships, Tuke writes, should not assume those relationships are permanent.

What Practitioners Should Take Away

The report explicitly rejects retreat as a response. Its prescription for the next phase of AI strategy is to treat it as a human problem before a procurement one. The firms closing the value gap fastest, Tuke writes, are not necessarily spending the most. They have named someone accountable for adoption landing in practice, measured what they actually want, and had the ownership conversation before it sours into resentment.

"The technology will keep improving regardless of what any single firm decides," Tuke concludes. "Whether your teams are ready to use it well, and trusted to remain answerable for the result, is still, for now, a choice."

The AI Futures Forum's prior reports tracked a hardening client mandate in September 2025, a warning about the collapsing training pipeline in "Broken Ladder" in January 2026, and a reckoning with the billable hour in May 2026. The full report is available as a free download. James Tuke is also the founder of Treat Digital and the author of "A Short Walk in AI," and advises firms and senior management on AI strategy.