Anthropic has announced ten new AI agents designed specifically for the financial sector. These are ready-made templates prepared for deployment in the real-world workflows of banks, investment funds, and insurance companies.
Each agent covers a different part of the financial world. Pitch Builder compiles company comparisons and drafts client presentations. Meeting Preparer prepares materials ahead of calls with counterparties. Earnings Reviewer reads annual reports and conference call transcripts, updates models, and flags changes relevant to the investment thesis. Model Builder creates and maintains financial models using regulatory filings, data sources, and analytical inputs. Market Researcher tracks developments across sectors and issuers, synthesizing news, filings, and brokerage research.
The second group of agents focuses on back-office operations. Valuation Reviewer compares valuations against the market and the methodology used. General Ledger Reconciler reconciles accounting entries and calculates net asset value. Month-End Closer carries out the closing checklist and prepares journal entries and reports. Statement Auditor checks financial statements for consistency and audit readiness. KYC Screener compiles client due diligence files, reviews source documents, and prepares escalations for compliance.
It is a logical step: financial work is largely repetitive, time-consuming, and governed by precisely defined rules. That makes it easier to automate than work in most other fields.
Claude in Excel and PowerPoint
The agents themselves are complemented by integration with Microsoft 365. Claude now works directly in Excel, PowerPoint, and Word, and will soon work in Outlook as well, through add-ins that can be installed in the tools you already use every day.
In practice, this means an analyst can start building a financial model in Excel. Claude assists throughout the process, organizing formulas across linked workbooks and running analyses. When it is time to create a presentation, the context is automatically transferred to PowerPoint. There is no need to explain anything again. The presentation is updated to reflect the revised figures. In Outlook, Claude then sorts the inbox, drafts replies in the user's voice, and organizes meetings.
How the agents work
Each template consists of three layers. Skills contain the instructions and domain expertise for the given task. Connectors provide controlled access to the data the agent needs. Sub-agents are additional Claude models that the main agent calls upon for specific subtasks, such as selecting comparable companies or reviewing the methodology.
The agents come in two forms. As a plugin in Claude Cowork or Claude Code, an agent runs alongside the analyst on their computer. As a Claude Managed Agent, it works autonomously on the platform, for example overnight or throughout an entire business case, with a complete log of every decision and tool access.
In both cases, a human remains in the decision-making loop. The agent does not send its outputs to the client on its own. People review, edit, and approve them.
Finance is Anthropic's second-largest source of revenue
The financial sector is Anthropic's second-largest source of enterprise revenue, after the technology industry. Forty percent of its fifty largest corporate customers come from finance. Citadel, Goldman Sachs, Citi, AIG, Carlyle, Mizuho, BNY, Walleye Capital, and Travelers are among the clients named by Anthropic.
Its move into the sector began last summer with the launch of Claude for Financial Services. Dario Amodei, CEO of Anthropic, said at an event in New York that the company's ability to monetize its technology and generate revenue is being constrained not by the performance of the models themselves, but by the speed at which they are rolled out across large organizations. It therefore needs partners that can help educate corporate clients and integrate the technology meaningfully into their operations.
Wall Street has been waiting for this, even though it has its own tools
Anthropic is entering a field that already has several strong leaders. Startups such as Rogo, valued at $2 billion and founded by former investment bankers, serve more than 250 clients with tools for presentations, research, and modeling. Hebbia, meanwhile, operates platforms that run concurrent queries across document collections and spreadsheets.
Rahul Rekhi, president of Rogo, says he is not overly concerned about growing competition. His tools are model-agnostic, so the better the models from Anthropic or other vendors become, the more Rogo can do for its clients. This means that the products most likely to succeed will be those that integrate easily with existing systems of record and operate within established risk management processes. Integration is not a minor detail, but a prerequisite for survival.
On the other side is OpenAI, which has also courted financial firms, with customers including BNY and BBVA. It is working on its own partnerships with private equity firms to distribute its tools. The race for the financial sector is certainly not limited to two companies.
Sources: forbes.com and businessinsider.com
