Ten ready-to-deploy AI agents just landed for bankers and your SaaS stack may not survive them
Anthropic released 10 AI agents built for investment banking and back-office financial work on Tuesday.
CEO Dario Amodei cited by Reuters warned that software companies failing to adapt to AI may “lose market value, go bankrupt, completely go bust.”
About 40 percent of Anthropic's top 50 customers are already financial institutions — its second-largest revenue segment after technology — with Goldman Sachs, Visa, Citi, AIG, and Citadel among them.
Anthropic's revenue grew "80x" on an annualised basis in the first quarter, Amodei told a New York audience alongside JPMorgan Chase CEO Jamie Dimon, Reuters reported.
Anthropic said the 10 agents cover research and client coverage work — pitchbook creation, meeting preparation, earnings review, financial modelling, and market research — as well as operations tasks including valuation review, general ledger reconciliation, month-end close, statement auditing, and KYC screening.
Each ships as a plugin in Claude Cowork and Claude Code, or as a cookbook for Claude Managed Agents, deployable within days rather than months.
As a managed agent, the template runs autonomously for work spanning a whole book of deals or a nightly schedule, with a full audit log where compliance and engineering teams can inspect every tool call and decision.
Users retain review and approval before work goes to a client or gets filed, Anthropic said.
Claude also now works across Microsoft Excel, PowerPoint, and Word via Microsoft 365 add-ins, with Outlook coming soon.
Context carries automatically between applications, so work started in Excel moves to PowerPoint without re-explaining anything, the company said.
Nicholas Lin, who leads Anthropic's financial services product work, told Reuters that AI model advances, hands-on customer support, and office software integrations were underpinning the rapid uptick in financial services business.
Anthropic announced new data connectors from Dun & Bradstreet, Fiscal AI, Financial Modeling Prep, Guidepoint, IBISWorld, SS&C IntraLinks, Third Bridge, and Verisk.
Moody's also launched an MCP app giving Claude access to credit ratings and data on more than 600m public and private companies for compliance, credit analysis, and business development.
Claude Opus 4.7 leads Vals AI's Finance Agent benchmark at 64.37 percent, the company said.
Amodei warned that Anthropic's Mythos model had uncovered tens of thousands of software vulnerabilities across industries, most of which remain unpatched.
CNBC reported that an earlier Claude model found roughly 20 vulnerabilities in Firefox; Mythos found nearly 300
Because Chinese AI rivals are “maybe six to 12 months” behind, there is roughly that long to address them before adversaries can exploit the same weaknesses, Amodei said.
Anthropic has limited Mythos to a few partner companies because of concerns about criminal or adversarial use.
Amodei called for AI regulation modelled on automotive safety standards.
Dimon said AI needs safety guardrails, likening it to a car without brakes. He called cybersecurity risks a “transitory period” but acknowledged AI's job impact was “a legitimate concern,” Reuters reported.
The launch comes as Anthropic and rival OpenAI race toward potential IPOs as early as this year, the Wall Street Journal reported.
Earlier this week, Anthropic announced a US$1.5bn joint venture with Wall Street firms to sell AI tools to private equity-backed companies, and partnered with Fidelity National Information Services to develop software helping banks flag financial crimes.
OpenAI has also been courting financial services firms, counting BNY and BBVA among its clients and working on a rival private-equity joint venture.
