Ema just raised $77M on the back of 50+ enterprise deals, including PwC and KPMG, with revenue growing 50-fold in two years. The story is simple: AI agent teams can now automate HR, IT, and finance processes that professional services firms traditionally staffed with bodies. For UK accountancy and financial services firms, this is both opportunity and threat. The opportunity is real — repetitive compliance work, data extraction, process scheduling, and routine client communications can move from junior staff to agents. The threat is equally real: if you don't control how and where agents operate, they will become a regulatory liability faster than they become a cost saving. The FCA's Consumer Duty (PS22/9) and the SRA's updated Code explicitly demand that firms remain accountable for work delivered to clients, regardless of whether a human or machine did it. Ema's customers may be automating internally. UK regulated firms automating client-facing work face a different standard entirely.

This story is part of a pattern we've been tracking for eighteen months: enterprise software companies are becoming less attractive as AI agents become more flexible. Why buy a static HR system when an agent can learn your specific processes, adapt to change, and handle exceptions? But there's a crucial distinction between what's happening in tech companies (where Ema's customers mostly are) and what can happen in law, insurance, accountancy, and financial services. Tech firms don't face FCA Consumer Duty obligations. They aren't subject to SRA ethical codes. Their work doesn't trigger PRA stress-testing requirements or ICO UK GDPR accountability frameworks. The same agent that optimises a tech company's invoice approval can create documented evidence of negligence or regulatory breach in a professional services context. Other AI vendors (Harvey in legal, Luminance in document review, Legora in compliance) have prospered precisely because they started from regulated problems, not general automation. Ema's strength is breadth and cost. That's not enough in regulated sectors.

Here's what Trovix believes, clearly: AI agents are genuinely useful for professional services, but only when they operate within a defined, auditable, and legally defensible scope. That means three things. First, agents should automate work that doesn't require professional judgment — data extraction, document classification, intake screening, regulatory filing checks. Second, that automation must be transparent and logged: a fee-earner must be able to see exactly what the agent did, why it did it, and what it flagged for human review. Third, the firm must be able to prove that humans made all material decisions. The firms succeeding with AI right now aren't the ones deploying agents to 'handle' client matters autonomously. They're the ones using agents to surface information faster so humans can decide better. Trovix's approach through Trovix Sift and Trovix Brief reflects this: we extract and structure, we don't replace. We flag, we don't decide. The vendors who promise full autonomy are selling a liability dressed as efficiency.

What should a mid-market accountancy firm or financial services house do about Ema's $77M and the wave of AI agents following it? Three concrete steps. First, audit where your real bottlenecks are — usually it's not complex judgment work, it's repetitive data handling, document management, and intake. Second, map those processes against your regulatory obligations under FCA, SRA, PRA, and ICO frameworks — if an agent fails at that task, what's the documented harm? Third, pilot AI carefully in low-risk zones with full auditability before touching anything client-facing. Ema will be useful to you, but not for what Ema is mainly used for today. You need AI that understands regulated environments, not AI that ignores them.

Source: TechCrunch