On May 5, 2026, Anthropic arrived on Wall Street with a very specific offer. The company unveiled ten agent templates built for financial services work, covering tasks that have historically consumed enormous amounts of analyst time: building pitchbooks, screening KYC files, reviewing earnings releases, reconciling general ledgers, and closing out each month. Each template is deployable the same day, either as a plugin inside Claude Cowork or Claude Code, or as a headless Managed Agent running autonomously via the Anthropic API. The announcement came alongside a press briefing in Lower Manhattan where CEO Dario Amodei appeared with JPMorgan Chase CEO Jamie Dimon, a pairing that signaled how seriously Anthropic is chasing institutional finance customers.
The timing was deliberate. Claude Opus 4.7, Anthropic's current flagship, had just posted the top score on Vals AI's Finance Agent benchmark, hitting 64.37% and outperforming every competing model on that evaluation. That number gave Anthropic a concrete credential to bring into conversations with investment banks, asset managers, and corporate finance teams who have started asking not just which AI model is smartest, but which one is most reliable on domain-specific work.
Ten Templates, One Architecture
The ten agents share a common internal structure. Each packages three components: skills, which are domain-specific instructions and knowledge for the task; connectors, which provide governed access to the data the agent needs; and subagents, additional Claude models called in for specialized sub-tasks such as comparables selection or methodology checks. Anthropic describes this layered approach as a way to let the main agent focus on orchestration while pushing precision work to models tuned for narrower jobs.
The full list spans two functional areas. On the investment-banking side: a pitch builder that assembles slide decks from company filings and comparable transactions; a meeting preparer that surfaces relevant deal history and counterparty intelligence; an earnings reviewer that reads through transcripts and highlights guidance changes; a model builder that populates financial models from live data sources; a market researcher that synthesizes competitive landscape information; and a valuation reviewer that checks DCF assumptions against current market comps. On the back-office side: a general ledger reconciler, a month-end closer, a statement auditor, and a KYC screener that processes client onboarding files against compliance requirements.
Anthropic Finance Agents: Key Facts
- Agent templates released10
- Vals AI Finance Agent benchmark score64.37% (industry leading)
- Microsoft 365 apps supportedExcel, PowerPoint, Word, Outlook (coming soon)
- Data partnerMoody's
- Deployment optionsClaude Cowork, Claude Code, Managed Agents API
- Functional areas coveredInvestment banking, back-office, compliance
Microsoft 365 and Moody's as Infrastructure
Two partnerships underpin the practical reach of these agents. The first is a deep integration with Microsoft 365. Claude now works directly inside Excel, PowerPoint, and Word through native add-ins, with Outlook support described as coming soon. Once installed, the add-ins carry context automatically between applications, so a financial model built in Excel can feed a deck being assembled in PowerPoint without the analyst re-explaining the structure in between. For a workflow like pitchbook construction, where analysts typically move data manually between a model and a presentation, the friction reduction is material.
The second is a data partnership with Moody's. Access to Moody's data gives the finance agents a governed, compliance-friendly source for credit ratings, company financials, and market intelligence, addressing one of the recurring objections institutional finance teams raise about AI tools: the question of where the model's information actually comes from and whether it can be audited. Connecting agents directly to a licensed data provider answers that question more cleanly than pointing to a general-purpose model trained on web-scale text.
"Claude Opus 4.7 is state-of-the-art on financial tasks and leads the industry on Vals AI's Finance Agent benchmark." Anthropic, finance agents announcement, May 2026
What Financial Firms Are Actually Buying
The finance agent push is part of a broader pattern. Anthropic has been methodically building vertical depth in industries where the underlying tasks are well-defined enough that agents can run reliably, and the labor cost of doing those tasks manually is high enough that buyers have a clear ROI story. Financial services checks both boxes: junior analyst work at investment banks and the compliance burden at retail banks and asset managers represent hundreds of billions of dollars in annual labor costs globally.
The reference-architecture approach, shipping reusable templates rather than bespoke deployments, also lowers the barrier for firms that want to experiment without committing to a full implementation project. A bank can deploy the KYC screener template, run it alongside existing workflows for a month, and measure its output against human reviewers before deciding whether to extend it to other use cases. That evaluation cycle is shorter than a traditional enterprise software procurement, which matters in a market where every major financial institution is running a parallel set of AI pilots and trying to consolidate around fewer vendors.
Anthropic's direct competitor in this space is a field that includes Harvey (which has focused on legal but has financial services ambitions), Bloomberg's AI offerings, and whatever internal tooling the large investment banks have built around models from OpenAI and Google. The finance benchmark score gives Anthropic a specific number to put in front of procurement teams, but the more durable advantage may be the connector infrastructure. Agents that already have governed access to Moody's data and live Microsoft 365 context are harder to replicate quickly than a model that scores well on a leaderboard.
For more on how Claude's enterprise integration strategy developed, see Claude's native integrations with Salesforce, SAP, and Microsoft 365. The finance agents also draw on the same Opus 4.7 model covered in last week's release notes. Anthropic's broader revenue trajectory, which forms the commercial backdrop for these vertical moves, is covered in our report on Anthropic's funding talks at a near-trillion-dollar valuation.
