Bloomberg's reporting on AI at top law firms confirms what we have been watching since 2024: the economics of legal service delivery are fracturing. Major firms are moving away from time-based billing because AI is making time-based pricing absurd. A legal research task that took four hours two years ago now takes 30 minutes with tools like Harvey or Luminance. Clients notice. They stop paying for those four hours. The alternative fee arrangements (AFAs) that were once edge-case offerings are becoming standard. For UK mid-market law firms, this is not an interesting trend—it is a competitive signal. The SRA's outcomes-focused regulation gives you freedom in billing models, but that freedom only helps if you move first. Firms that delay are ceding margin to competitors who have already restructured their cost base around AI-assisted delivery.

What Bloomberg's story reveals is that AI adoption at scale has moved past the research-and-development phase and into operational reality. Hiring, billing, and the substance of legal work itself are all being redesigned simultaneously. This is not isolated innovation—it is systemic transformation. The traditional leverage model (junior lawyers doing commodity work to subsidise partnership income) is collapsing because AI does commodity work better and cheaper. Firms cannot hire their way out of this. They must restructure around outcomes, specialisation, and genuine client value rather than hours worked. This shift favours firms with clear AI governance frameworks, because without them, the cost savings disappear into chaos: unbacked outputs, compliance risk, poor client experience. The FCA Consumer Duty (PS22/9) and the SRA's emerging AI guidance both require firms to understand and govern their AI systems—not just use them.

We are critical of vendors (including some very well-funded ones) who sell AI tools as cost-cutting instruments without addressing governance. They are selling you a margin trap. Harvey and Legora have built powerful models, but they are research and generation engines. They do what they were designed to do—generate text, find patterns—at scale and speed. What they do not do is sit at the governance layer. A firm deploying Harvey across 200 lawyers without auditable workflows, without clear ownership of output quality, without documented AI decision-making, will hit the SRA's desk within two years. We have seen it. Trovix's approach starts with Trovix Audit, which means before you scale AI use, you build the compliance and governance layer. Then you add Trovix Aria for your fee-earners—a knowledge assistant that works within your governance envelope—and Trovix Sift for document work, so you extract and process data under audit. The margin gains are real, but they only stick if you are not managing regulatory risk in parallel.

If you lead a mid-market law firm, insurer, accountancy practice or financial services outfit, you have a 12-month window. Act now on three things: (1) Audit your current AI use—what is being deployed, by whom, under what controls? (2) Restructure at least one high-volume fee-earner workflow (billing review, due diligence, compliance checks, contract extraction) around AI plus governance, and measure the margin and quality impact. (3) Start hard conversations with your clients about alternative fee arrangements, outcome-based pricing, or hybrid models. Firms that wait for AI to become 'safe' or 'standard' will wait too long. Firms that deploy AI without governance will pay for it in regulatory attention and reputational cost. The firms that win are the ones moving now with both tools and controls in place.

Source: Bloomberg News