Anthropic's 10 new financial services agents—announced in May—have already spooked FactSet and Morningstar investors. The tools draft pitch decks, review financial statements, and escalate compliance cases. For UK mid-market firms in banking, insurance, asset management and accountancy, this looks like another wave of direct competition for work you currently pay specialists to do. But this is exactly where the framing matters. These are useful automation tools. They are not—and cannot yet be—genuine replacements for regulated financial decision-making. The FCA's Consumer Duty PS22/9 and its ongoing emphasis on governance and competent staffing make this distinction regulatory, not rhetorical.

The pattern is now undeniable. First came document review tools. Then came contract analysis platforms like Harvey and Luminance, which solved narrow, well-defined problems exceptionally well. Now we are watching the industry move toward 'agents'—autonomous systems that chain together multiple capabilities and make decisions with minimal human intervention. Anthropic's approach mirrors what we are seeing across the sector: vertical solutions pitched at professional services. Each is competent at its narrow task. But here is what the market is not yet admitting: most real compliance, investment, and legal work is not narrow. It is contextual, regulatory-dependent, and requires ongoing human judgment. The FactSet and Morningstar decline tells you investors are beginning to wonder if agents can really do what vendors claim.

We have a blunt view at Trovix: AI agents work best when they augment, not replace, human expertise—and when they are built to integrate with your actual workflow, not sold as standalone tools. Anthropic's new agents are general-purpose. They will hallucinate on UK regulatory specifics. They will miss the nuances between FCA, PRA, and SRA obligations that determine whether a decision is compliant or reckless. Compare this to how Trovix Aria works: it is built as a knowledge assistant for fee-earners and compliance teams, trained on your firm's own precedent and regulatory framework, embedded in your workflow, with audit trails that satisfy the ICO UK GDPR and ISO 42001 governance expectations the FRC now expects of professional firms. That is not anti-agent. It is pro-competence. Standalone agents, without the infrastructure of your own compliance knowledge and human oversight, are a liability dressed as efficiency.

Here is what to do: Do not panic. Do not assume that because Anthropic has released financial agents you must immediately adopt them or lose competitive ground. Instead, audit where your firm actually wastes time and where it actually needs judgment. Pitch deck generation? Agents handle that. Evaluating whether a transaction structure meets PRA SS1/23 on operational resilience or the EU AI Act's emerging transparency rules? You need human expertise plus AI that understands your regulatory context. Invest in tools that lock down compliance risk—Trovix Watch for regulatory monitoring, Trovix Sift for document intelligence where scale matters—rather than generic agents that do many things without mastering any.

Source: Bloomberg News