Forbes reports that $2.1 billion flowed into legal-tech startups in H1 2026, with Google, Harvey, and newer entrants racing to automate contracts and incorporation. This is important to UK mid-market law firms, insurers, and accountancy practices for one reason: it signals that document automation and intake AI are now table stakes, not differentiators. But here is what the $2 billion story does not tell you: it does not tell you how to implement any of these tools without exposing your firm to regulatory risk under the SRA Code, FCA Consumer Duty PS22/9, or the emerging EU AI Act compliance obligations that UK regulators will mirror. The money went into product velocity. It did not go into solving governance.

This is part of a larger pattern we have been watching for eighteen months. Venture capital funds AI vendors who build clever models, not AI vendors who solve the hardest problem in professional services: proving to your regulator that you remain accountable for work an AI system has done. The tech press celebrates the product announcements. The in-house counsel and compliance teams at law firms sit in meetings asking the uncomfortable question: if this AI makes a mistake in a contract review, and a client is harmed, who is liable and what audit trail do we have? Most vendors have an unsatisfying answer. That is because most vendors are not built by people who have sat through SRA investigations or FCA enforcement reviews.

Our view at Trovix is this: document automation and intake AI are genuinely useful. But they are only useful if they are wrapped in governance that makes them auditable. That means you need three things that most point solutions—including Harvey and Luminance—do not offer as standard: real-time explainability of every AI decision (not just accuracy metrics), a compliance dashboard that speaks to your regulator's actual concerns under ISA UK and ISO 42001 frameworks, and a human-in-the-loop workflow that is not a checkbox but a genuine control. We built Trovix Sift for document intelligence precisely because law firms and accountancy practices asked us: how do we automate without losing accountability? And we built Trovix Audit because the real cost of AI implementation is not the software. It is the regulatory risk if you cannot explain what the software did and why.

Here is what you should do Monday morning: stop asking your AI vendor whether the model is accurate. Start asking them whether they can prove to the SRA, FCA, or PRA that you remain the accountable person under the Duty of Care rules and Consumer Duty regime. Ask for a working audit trail, not a promise of one. Ask them how they handle conflicts between speed and explainability—because there is a trade-off and anyone who says there is not is selling something. Then run a pilot on your riskiest, most repeatable workflow: estate administration, contract schedules, client intake forms. Measure not just time saved, but your ability to justify every decision to a regulator if things go wrong. The $2 billion went to the vendors who moved fastest. Your job is to be the firm that moved smartest.

Source: Forbes