Anthropic released AI agents for finance automation with 10 ready-to-run templates designed to automate tedious Wall Street finance tasks. The templates are available as plugins in Claude Cowork and Claude Code and as cookbooks for Claude Managed Agents; each includes skills, connectors, and subagents to work across Excel, PowerPoint, Word and Outlook with context carried over between apps. Anthropic noted firms can have Claude working on real financial tasks within days rather than months, enabling deployments for pitch building, KYC screening and month-end close processes.
Anthropic’s AI agent templates are organized into two primary categories: research and client coverage, and finance and operations. Within the research and client coverage category, the templates include a pitch builder, a meeting preparer, and an earnings reviewer. These tools are designed to streamline the preparation of comprehensive pitchbooks and conduct thorough reviews of earnings reports. In the finance and operations category, templates are available for general ledger reconciliation, month-end close checklists, financial statement audits, and know-your-customer (KYC) file screening. Each template is equipped with skills, connectors, and subagents specifically tailored to execute these tasks efficiently, providing firms with the resources needed to enhance productivity and accuracy in financial processes.
Anthropic’s Claude AI agents offer robust integration capabilities, working seamlessly across Microsoft Excel, PowerPoint, Word, and Outlook. This integration allows for retention of context across different applications, enhancing efficiency and workflow continuity. The latest connectors enrich Claude’s capabilities by sourcing data from Dun & Bradstreet, Fiscal AI, Guidepoint, IBISWorld, SS&C IntraLinks, Third Bridge, and Verisk, allowing firms to integrate extensive external datasets into their financial analyses. Additionally, the Moody’s MCP app plays a crucial role by providing proprietary credit ratings and data on over 600 million public and private companies, broadening the scope of financial intelligence available to users. Performance-wise, Claude Opus 4.7 leads the Vals AI Finance Agent benchmark, attaining a score of 64.37%, indicating its superiority in handling financial tasks among similar AI solutions.
Anthropic noted that users remain in the loop for review and approval before AI-generated output is sent to clients or filings. Mozilla discovered vulnerabilities in Anthropic’s Mythos model, and those vulnerabilities were subsequently patched in Firefox. These points — user oversight before external distribution, and the discovery and subsequent patching of Mythos vulnerabilities — were reported alongside Anthropic’s deployment of AI agents for finance automation.
Anthropic delivered 10 ready-to-run templates and integrations that enable Claude to operate across Excel, PowerPoint, Word and Outlook, with templates available as plugins in Claude Cowork and Claude Code and as cookbooks for Claude Managed Agents. These ready-to-deploy templates, each including skills, connectors and subagents, accelerate deployment so firms can have Claude working on real financial tasks within days rather than months, while keeping users in the loop for review and approval before outputs reach clients or filings.
