Forbes' headline captures the paradox perfectly: $2.1 billion of capital has poured into legal-tech startups that promise to slash document costs by automating drafting and letting licensed attorneys rubber-stamp the output. That's real money solving a real problem. But for UK mid-market law firms, accountancy practices, insurers and financial services firms operating under the SRA Code, FCA Consumer Duty PS22/9, and PRA SS1/23, this investment trend should trigger a hard question: if AI can generate compliant documents with 95% accuracy, who carries the risk when it doesn't? The story rightly notes that human judgment remains critical. The problem is that most of the $2.1 billion is being spent on tools that treat attorney review as a checkbox, not a redesign of how expertise gets deployed.

What we're seeing is the predictable pattern of venture capital finding a bottleneck—expensive human time spent on document drafting—and funding companies to automate it away. Harvey, Legora, Luminance and the rest have built impressive products. But the industry is converging on a dangerous assumption: that licensed oversight plus AI output equals solved risk. It doesn't. The EU AI Act's risk-based framework and the ICO's emerging guidance on generative AI make clear that regulated firms cannot treat AI-generated content as inherently lower-risk because a human saw it afterwards. A solicitor signing off a lease drafted by an LLM that hallucinated a liability clause hasn't reduced negligence exposure—they've distributed it. The real gap this $2.1 billion wave is not filling is how to actually understand what an AI system has done, where it might fail, and whether the firm's insurance and governance structures can absorb that risk.

Trovix's view: AI should augment judgment, not substitute for it disguised as oversight. That means firms need two things the current vendor landscape mostly ignores. First, genuine document intelligence—the ability to extract, flag and explain what an AI system has actually changed or generated, not just approve it faster. Trovix Sift was built specifically for this; it surfaces what's actually in documents rather than hiding automation behind a sign-off button. Second, firms need live awareness of how regulations around AI governance are shifting—because by 2027, the SRA will have tightened its guidance on AI use in reserved legal work, and the FCA is already signalling stricter rules on AI-assisted financial advice. Trovix Watch gives firms that visibility, so they're not caught flat-footed when the rule changes. The firms winning with AI are the ones treating it as a transparency tool first, not a cost-cutting tool first.

For a 50-partner law firm, a mid-size accountancy practice or a specialist insurance broker, the decision is not whether to adopt AI—it's how to adopt it without creating downstream liability. Run a genuine pilot. Measure not just speed but what changes, why, and whether your E&O insurance covers AI-assisted work products. Document the control process, not just the output. Talk to your regulator informally—the SRA, FCA and PRA are genuinely interested in how firms are implementing this responsibly. And be honest: if your chosen AI tool can't explain what it changed or why, you're outsourcing risk, not managing it.

Source: Forbes