In May 2025, we published our first benchmark of how UK professional services firms were using AI. By January 2026, our “Broken Ladder” Green Paper warned that automating routine tasks was quietly destroying the training ground for the next generation of professionals. In May 2026, our most comprehensive report yet, “AI, Human Capital and the Future of Professional Services,” mapped the full scale of the transition: the breakdown of the billable hour, the rise of “NewMod” competitors, and a profession caught between hype and reality.

Four months on, the industry has caught up with what we were describing. Everyone is now talking about the “AI value gap” – Thomson Reuters’ finding that 91% of professionals say their organisations are falling short of what AI could deliver. What nobody has properly answered is why the gap persists even in firms that have invested seriously, hired well, and licensed the best tools on the market.

Our new report, “The AI Value Gap: The Human Economics of Stalled AI in the Professional Services Sector“, sets out to answer exactly that.

It’s Not a Technology Problem. It’s a Confidence Problem.

The usual explanations – immature tooling, thin governance, patchy training – are real, but they don’t explain a gap this size. Our research identifies four human failure modes sitting underneath the headline statistics: leadership paralysis, where firms recognise AI’s importance but can’t convert conviction into commitment; a trust deficit that shows up as unsanctioned “shadow AI” rather than open engagement; an identity threat, as the tasks bound up with professional mastery are the first to be automated; and incentive structures that quietly punish the very efficiency gains firms claim to want.

The Ownership Question

We also surface a dimension the sector has barely started to discuss: as firms train AI systems on their own partners’ accumulated judgement, who actually owns the resulting asset? It’s no coincidence that Kirkland & Ellis’s $500 million platform bet was read by commentators as being about ownership and control, not efficiency – or that we’re now seeing firms explore spinning out their own AI capability as a separate commercial entity.

A New Kind of Competitor

The competitive threat here isn’t confined to well-funded “NewMod” start-ups any more. In 2026, Anthropic’s launch of Claude for Legal signalled that foundation labs themselves are moving directly into the application layer built on top of them – a genuine Trojan Horse risk that even firms comfortable with their current AI vendor need to reckon with. OpenAI has just followed suit with their own offering: Astra for Law.

Where Does Your Firm Actually Sit?

The report closes with a self-assessment tool, built entirely from published survey data rather than speculation about named competitors, so firm leaders can work out honestly where they sit before deciding what to fix first.

Request Your Copy of the Full Report here