Anthropic's Financial Services Push Is Impressive. It's Not What Our AI Advisory Forecast Is Actually Measuring.
TexTak places 35% odds on a major bank launching an AI-only financial advisory product for retail customers — and today's Anthropic news is being widely read as confirmation that this threshold is approaching. We think that reading conflates two very different things. Anthropic's financial services expansion is genuinely significant. It is also almost entirely pointed at institutional workflows, not retail advisory — and that distinction is the entire forecast.
Let's be precise about what Anthropic actually announced. Ten pre-built financial agents targeting pitchbooks, audit statements, and credit memos. Integrations with FactSet, S&P Capital IQ, MSCI, PitchBook, and institutional data warehouses. FIS deploying an AML investigation agent compressing hours-long workflows to minutes. SMBC deploying an automatic proposal generation system for wholesale banking corporate advisory. Claude for Outlook in beta for analysts toggling between spreadsheets and email. This is sophisticated, production-grade institutional AI — and it represents genuine enterprise deployment, not pilot theater. But every single use case targets analysts, bankers, and compliance officers. None of it is pointed at the retail customer relationship our forecast is actually measuring.
Our 35% reflects a specific threshold: an LLM-powered advisory product from a major bank providing personalized financial advice — not portfolio allocation via algorithm, which robo-advisors have done for a decade — to retail customers, with the bank's brand on it. The barrier isn't technical capability; Anthropic has now demonstrated the underlying capability is clearly there. The barrier is SEC and FINRA regulatory requirements for investment advice, fiduciary liability exposure, and the reputational catastrophe risk of an AI system giving a retail customer materially bad advice that makes the front page. JPMorgan and Morgan Stanley deploying internal AI tools is proximate evidence — it proves the capability is entering the institution. It does not prove the institution is ready to face a retail customer without a human advisor in the loop.
Honestly, the part of our thesis that today's news most pressures is the timeline assumption embedded in 35%. Anthropic's expansion velocity in financial services is faster than we modeled six months ago. The Goldman Sachs, Visa, Citi, and AIG adoption pattern suggests these institutions are moving through institutional comfort zones faster than their historical conservatism would predict. If the wholesale banking use cases currently being deployed generate clean track records through 2025 and early 2026, the internal risk calculus for a retail product could shift faster than our probability implies. We're watching whether any major bank files for a no-action letter with the SEC around an AI advisory product — that would be the specific regulatory signal that retail deployment is being seriously planned, not just discussed.
What would move us above 50%: SEC or FINRA issues guidance specifically enabling AI-only advisory relationships, removing the regulatory ambiguity that currently makes banks' legal teams veto the product concept. What would drop us below 20%: a high-profile AI advisory failure at the institutional level — an AML false positive causing regulatory action, or an AI-generated credit memo error causing material loss — that triggers risk committee freezes on customer-facing AI expansion.