Bloomberg's reporting on AI at Freshfields and other top-tier practices confirms what mid-market UK firms are beginning to understand: the billable hour — the economic engine of law for fifty years — is structurally incompatible with AI-assisted work. When a fee-earner completes a document review in four hours using AI that would have taken twenty hours manually, firms face an immediate choice: bill less, release capacity to other work, or charge for efficiency gains. The SRA Code of Conduct 2019 permits outcome-based and fixed fees, yet most practices still anchor everything to time. That's no longer sustainable. This isn't disruption happening to someone else's practice. It's happening to yours right now.

What Bloomberg's story reveals is the acceleration of a three-part shift already underway in UK regulated firms. First, the tools themselves — Harvey, Luminance, Copilot, and homegrown solutions — are moving beyond research assistants into core matter delivery. Second, hiring is shifting from junior fee-earners doing high-volume work towards specialists who can manage AI systems, review AI output, and handle complex judgment calls. Third, pricing is fragmenting. Firms charging by the hour for AI-assisted work are cannibalizing their own margins. Firms charging by the task or outcome are capturing the value they've created. The gap between the two approaches is widening fast. This pattern will accelerate as AI tools become more capable and as clients — especially in-house counsel managing legal budgets under FCA Consumer Duty PS22/9 standards — demand better pricing transparency.

Here's where honest implementation differs from the marketing noise. Tools that promise 'end-to-end' legal work automation (Harvey, for instance) work brilliantly on narrow, repetitive tasks where outputs are easily validated. They fail badly when judgment, context, and regulatory nuance matter — which is most of what UK legal, insurance, and accountancy firms actually do. The smarter approach, which we've built into Trovix Aria and Trovix Sift, is to treat AI as intelligence amplification, not replacement. Your fee-earners need speed on facts, pattern recognition on large document sets, and reliable extraction of structured data. They need to spend thinking time on advice, strategy, and client relationships — the work that actually justifies fees. Tools that do retrieval-augmented generation properly (Aria) let practitioners query institutional knowledge like having a senior colleague instantly available. Tools that extract and classify documents accurately (Sift) turn manual triage from a cost centre into a margin generator. Neither promises to turn a junior into a partner. Both make good people better.

What should a mid-market firm do on Monday morning? Stop asking 'which AI platform should we buy' and start asking 'which workflow creates the most fee-earner time back for judgment work?' Map your highest-volume, lowest-judgment tasks first — document review, billing code classification, regulatory change spotting (Trovix Watch handles that specifically), intake data extraction (Trovix Brief automates this). Measure current time cost and error rate. Then pilot AI on one matter. Measure time saved, quality, and how your fee-earners actually behaved with the tool — did they trust it? Did they skip validation? Did they use the time freed for better work, or just bill less? That data will tell you whether you're on a path to better economics or just cheaper mediocrity. And when you redesign pricing, be explicit about it with clients. Transparency about how AI changes your delivery model and their costs is now a competitive advantage, not a liability.

Source: Bloomberg News