Anthropic released 10 pre-built AI agent templates targeting the most labor-intensive workflows in banking and financial services, the company announced at its invite-only financial services briefing in New York on Tuesday. The templates ship alongside Claude Opus 4.7, which leads the Vals AI Finance Agent benchmark at 64.37%.

This is the most aggressive vertical-specific agent deployment from any frontier lab to date. Rather than selling a general-purpose model and letting banks figure out integration, Anthropic is packaging domain knowledge, data connectors, and multi-agent architectures into templates that teams can deploy in days.

The 10 Agent Templates

Anthropic split the templates into two categories matching real financial workflows, per the company blog:

Research and Client Coverage: Pitch Builder generates target lists, runs comparable analyses, and drafts pitchbooks. Meeting Preparer assembles client and counterparty briefs. Earnings Reviewer reads transcripts and filings, updates financial models, and flags material changes. Model Builder creates and maintains complex models from multiple data sources. Market Researcher monitors sectors, synthesizes news and broker research, and flags items for credit and risk review.

Finance and Operations: Valuation Reviewer cross-checks valuations against comparables and internal guidelines. General Ledger Reconciler manages account reconciliations and NAV calculations. Month-End Closer runs close checklists, prepares journal entries, and produces close reports. Statement Auditor reviews financials for accuracy and audit readiness. KYC Screener compiles entity documentation, examines source materials, and packages escalations for compliance review.

Each template is a reference architecture packaging three components: skills (domain instructions and knowledge), connectors (governed data access), and subagents (additional Claude models called for specific subtasks like comparables selection or methodology checks). Firms can adapt them to their own modeling conventions, risk policies, and approval chains, according to Anthropic.

Deployment Paths

The templates are available two ways. As plugins in Claude Cowork and Claude Code, they run alongside analysts using existing desktop software. As cookbooks for Claude Managed Agents, they run autonomously on the Claude Platform with long-running sessions, per-tool permissions, managed credential vaults, and full audit logs.

Anthropic also launched Claude add-ins for Microsoft 365 covering Excel, PowerPoint, and Word, with Outlook coming soon. Context carries between applications, so a model built in Excel can flow into a deck in PowerPoint without re-explanation, per the company.

Data Ecosystem Expansion

The agents connect to existing platforms including FactSet, S&P Capital IQ, MSCI, PitchBook, Morningstar, Chronograph, LSEG, and Daloopa. New connectors announced at the event include Moody’s (covering 600 million+ companies), Dun & Bradstreet, Guidepoint (100,000+ expert interview transcripts), SS&C IntraLinks (deal data rooms), Third Bridge (primary-source expert interviews), IBISWorld, Fiscal AI, Financial Modeling Prep, and Verisk, according to Anthropic.

Early Deployments

Fidelity National Information Services (FIS) is using the agents for anti-money laundering reviews, with expansion planned into credit decisions and fraud detection. Private equity firm Carlyle has deployed Claude across investment analysis, operations, and portfolio oversight, Fortune reported.

Anthropic’s Chief Commercial Officer Paul Smith described a “staircase of autonomy” at the event, outlining progression from basic research support to fully autonomous work at senior analyst level, according to Fortune.

The Verticalization Play

The template approach marks a strategic shift from selling model access to selling workflow automation. Disruption Banking noted that Anthropic is “constructing the infrastructure, deployment systems, and high-level relationships needed to become the foundational operating layer for Wall Street.” The company is competing directly with OpenAI to prove its technology can handle high-value work beyond coding, with both companies seeking to expand revenue ahead of expected IPOs.

For financial services firms evaluating the templates, the key variable is how much customization the reference architectures actually require. Pre-built agents that need months of configuration to match internal policies are demos, not products. Agents that deploy in days and integrate with existing data pipelines are infrastructure. Anthropic is betting on the latter.