Anthropic’s AI Agents Are Here to Transform Day-to-Day Financial Services and Insurance

AI agents are no longer just a promise — they’ve become real working tools in the financial market.

If you’ve ever spent hours building a pitchbook from scratch, reconciling accounts at the end of the month, or reviewing KYC files one by one, you know exactly what I’m talking about. This kind of repetitive, time-consuming work doesn’t require creativity, but it demands total focus — and any distraction can be costly. In the financial sector, typos, outdated data, or inconsistencies in reports have real consequences, from regulatory fines to misguided investment decisions.

This type of work eats up valuable time that could be spent on analysis, strategy, and client relationships. Qualified professionals end up stuck in operational tasks that could — and now can — be automated with intelligence and precision.

Anthropic saw this problem up close and decided to do something concrete about it. The company just launched ten ready-to-use agent templates built specifically for the heaviest tasks in financial services and insurance — from building pitchbooks and financial models to closing monthly balance sheets and screening KYC files.

And it doesn’t stop there. Alongside the agents come new integrations with the Microsoft 365 suite and a significant expansion of the financial data ecosystem, with connectors from partners like Moody’s, Dun & Bradstreet, Verisk, and IBISWorld, among others. All of this runs on Claude Opus 4.7, the model currently leading the Vals AI Finance Agent benchmark with a 64.37% performance score — the best in the industry for financial tasks.

In the following sections, we’ll break down everything that was launched, how it works in practice, and what the major financial institutions already using it have to say about it. 🚀

The Ten Agents Changing the Game in Financial Services

Anthropic didn’t show up with a generic solution. The ten templates were designed around the real pain points of people working in the financial market day in and day out, and each one covers a specific workflow that typically drains precious hours from teams.

Each agent template is essentially a reference architecture that bundles three core components: skills (instructions and domain knowledge for the task), connectors (governed access to the data the task needs), and subagents (additional Claude models triggered by the main agent for specific subtasks, like comparable selection or methodology checks). Institutions can customize any of them to fit their own modeling conventions, risk policies, and approval workflows.

These agents are divided into two main categories:

Client Research and Coverage

• Pitch Builder — creates target lists, runs comparable analyses, and assembles complete pitchbooks for client meetings.
• Meeting Preparer — pulls together client and counterparty briefings ahead of calls and meetings.
• Earnings Reviewer — reads earnings transcripts and regulatory filings, updates models, and flags relevant changes to the investment thesis.
• Model Builder — creates and maintains financial models from balance sheets, data feeds, and analyst inputs.
• Market Researcher — tracks sector and issuer developments, synthesizes news, regulatory filings, and broker research, and flags items for credit and risk review.

Finance and Operations

• Valuation Reviewer — checks valuations against comparables, methodology, and the firm’s review standards.
• General Ledger Reconciler — reconciles general ledger accounts and runs net asset value calculations against the books of record.
• Month-End Closer — runs through the closing checklist, prepares journal entries, and produces close-out reports.
• Statement Auditor — reviews financial statements for consistency, completeness, and audit readiness.
• KYC Screener — gathers entity files, reviews source documents, and packages escalations for compliance review.

The core idea is simple: instead of building an agent from scratch, financial institutions get a functional, already-optimized foundation that can be customized to fit each operation’s needs. This drastically cuts implementation time and puts technology teams in a much more comfortable position to deliver quick results.

Two Ways to Put the Agents to Work

There are two paths for using these templates. As a plugin in Claude Cowork or Claude Code, the agent runs alongside the analyst, using the software they already have on their desktop. Hand the Pitch Builder agent a list of targets, and it can deliver a comparables model in Excel, a fully assembled pitchbook in PowerPoint, and a cover note ready to go in Outlook.

As a Claude Managed Agent, the same template runs autonomously on the Claude Platform, ideal for jobs that span an entire book of business or an overnight schedule. The cookbooks provide long-running sessions capable of operating during multi-hour deal closings, per-tool permissions, managed credential vaults, and a complete audit log in the Claude Console, where compliance and engineering teams can inspect every tool call and every decision the agent made.

In both scenarios, users remain in control — reviewing, iterating, and approving Claude’s work before it gets sent to a client, filed, or executed.

Microsoft 365 Integration: Excel, PowerPoint, Word, and Outlook

One of the most practical features of this launch is the direct integration with Microsoft 365 through dedicated add-ins for Excel, PowerPoint, Word, and Outlook (the latter coming soon). For anyone working at banks, asset managers, and brokerages, this is no small detail.

Most financial operations still run through Excel, Word, and PowerPoint — and having Claude operate directly inside these tools, without needing a separate platform, completely changes the adoption dynamic. The analyst doesn’t need to learn a new system, doesn’t need to export data to another tool, doesn’t need to change the workflow they already know. The agent goes where the work happens, not the other way around.

In Outlook, Claude works like a chief of staff, triaging the inbox, organizing meetings, and drafting replies in the user’s own tone. In Excel, it builds financial models from balance sheets and data feeds, audits formulas across linked workbooks, and runs sensitivity analyses. In PowerPoint, it assembles presentations that auto-update when the underlying numbers change. In Word, it edits credit memos based on the firm’s own templates.

The most interesting part is that Claude carries its knowledge and context across all four platforms. An analyst who started a model in Excel doesn’t need to re-explain anything when that work moves to PowerPoint. This cross-application context continuity is the kind of functionality that eliminates real friction in daily workflows.

In Claude Cowork, users can also assign tasks to Claude from anywhere — by text or by voice — using a feature called Dispatch. Claude can keep working on the analyst’s local files while they’re away from their desk, with the finished work ready for review when they return. 💻

A Broader Data Ecosystem for Financial Services

AI agents are only as good as the data and context they can access. And Anthropic knows it. Claude already connects to dozens of market data platforms, research sources, and internal enterprise systems at financial firms — including S&P Capital IQ, MSCI, PitchBook, Morningstar, Chronograph, LSEG, and Daloopa — plus firms’ own data warehouses, research repositories, and CRMs, all under governed access controls.

Now, new connectors and a partner MCP app are being added to the ecosystem:

New Connectors

• Dun & Bradstreet — provides the global standard for verified business identity and helps companies connect systems of record and scale AI-enabled workflows.
• Fiscal AI — extends real-time fundamentals coverage for public equities, deepening research and benchmarking.
• Financial Modeling Prep — offers real-time quotes, fundamentals, financial statements, regulatory filings, and transcripts for stocks, ETFs, crypto, forex, and commodities.
• Guidepoint — searches over 100,000 compliance-reviewed expert interview transcripts and provides verbatim excerpts linked to the source.
• IBISWorld — tracks industry revenues, financial ratios, risk scores, cost structures, and forecasts across thousands of sectors.
• SS&C IntraLinks — gives Claude access to DealCentre data rooms for document search, Q&A due diligence, and deal activity tracking.
• Third Bridge — grants Claude access to primary-source expert interviews on companies, industries, and value chains.
• Verisk — provides property, casualty, and specialty insurance data for underwriting, claims, and risk analysis.

Moody’s MCP App

Beyond the connectors, Moody’s launched an MCP app that brings proprietary credit ratings and data on more than 600 million public and private companies for use in compliance, credit analysis, and business development. This app goes beyond a simple connector because it incorporates Moody’s own tools directly inside Claude, creating an integrated user experience that eliminates the need to consult external platforms in parallel.

When these data streams combine inside an intelligent agent, the level of analytical depth a small team can deliver increases considerably. And the architecture was built to scale — Anthropic has signaled that the partner ecosystem will continue to expand, which means institutions that adopt the agents now are investing in a platform that’s only going to get more powerful over time.

Claude Opus 4.7 Leading the Financial Benchmarks

Behind all these agents sits Claude Opus 4.7, and the numbers it puts up on the specialized Vals AI Finance Agent benchmark are no accident. Reaching 64.37% performance on this kind of evaluation requires the model to not only process natural language but also reason about financial data, follow complex business logic, and produce outputs that meet the sector’s rigorous standards. This specific benchmark tests exactly the types of tasks the agents were designed to execute, which makes the result even more relevant for anyone evaluating adoption in real production environments.

The model’s superior performance on financial tasks reflects a deliberate effort by Anthropic to specialize Claude for specific verticals. Rather than betting solely on an ever-larger generalist model, the company invested in fine-tuning that makes the model more precise and reliable in contexts where errors carry real weight. In the financial market, a wrong number in a regulatory report or a poorly calibrated projection can generate serious consequences — and that’s why the combination of technical precision and contextual reasoning in Opus 4.7 is so valued by the teams already using the tool.

What the Institutions Already Using Claude Have to Say

There’s no shortage of testimonials from big names in the financial market already seeing concrete results with Claude. And it’s worth listing some of the most relevant ones here, because they show the diversity of applications and the depth of adoption.

Citadel highlights that their investment professionals live immersed in data and analytical models, and that Claude for Excel meets them exactly where they are. Analysts are using the tool to build and update coverage models, separate signal from noise, and stress-test their analyses — all with a significant jump in efficiency.

FIS, which sits at the center of how money moves for thousands of financial institutions around the world, chose Anthropic to build agents that compress anti-money laundering investigations from days to minutes, with credit decisioning, fraud prevention, and deposit retention agents coming next.

BNY talks about giving processes new digital workers that work cases end-to-end with Eliza and Claude. Carlyle adopted Claude as a central part of its AI tech stack, citing strong coding capabilities, agentic reasoning, and continuous advances in both models and core features.

Mizuho reports that Claude compresses and enhances pre-meeting work, turning prep time into idea time, with faster workflows, richer insights, and use cases they hadn’t even anticipated.

Tools we use daily

Translation

Text Inspection & Clipping

Social Media

Search & Research

Recruitment

Programming

Productivity & Organization

Operations

Marketing & Sales

Finance & Investment

Education

Design & Media

Data & Analytics

Content & Writing

Automation

Travelers has observed significantly elevated levels of engineering excellence and real productivity improvements since introducing custom Claude assistants and Claude Code.

And perhaps the most impressive number comes from Walleye Capital: 100% of employees at the 400-person hedge fund use Claude Code. As the firm itself puts it, this adoption reflects an AI-first mindset, where everyone is constantly rethinking how they work.

Hg highlights Claude for Excel with Opus as a significant leap forward, especially in due diligence and financial modeling, automating complex analyses with minimal prompting needed. Morningstar and PitchBook reinforce that trust in AI starts with the data behind it, and their decades of independent intelligence deliver answers that aren’t just faster but better. And FactSet reports that adopting Claude Code across its entire engineering organization is accelerating the speed at which new capabilities are delivered to clients. 🏦

What Actually Changes for Analysis and Compliance Teams

For analysis teams, the most immediate impact is the ability to scale work without proportionally scaling the team. With agents handling the initial data collection, standardization, and organization, analysts get to focus on the layer that truly requires human judgment: interpreting the numbers, identifying trends, and building the narrative that will support strategic decisions. This doesn’t eliminate jobs — it reorganizes where human talent is applied, and that shift tends to be positive both for the quality of work delivered and for team satisfaction, as people stop being hostages to repetitive, draining tasks.

For compliance and KYC operations, the difference is even more visible. The volume of documentation that financial institutions need to review grows every year, driven by stricter regulations and an expanding client base. Having a Claude agent that can review documents, identify inconsistencies, and flag cases that need human attention drastically reduces new client onboarding time and lowers the risk of regulatory failures that would slip through the cracks in manual reviews done under time pressure. The integration with partner data systems like Dun & Bradstreet makes this process even more robust, because the agent can cross-reference information from trusted external sources in real time, without relying on parallel manual lookups by analysts.

In the long run, what these agents represent is a shift in how financial institutions think about automation. It’s no longer about automating isolated tasks with rigid scripts that break every time a format or regulation changes. Claude agents are flexible enough to adapt to variations in data and workflows, which makes them far more resilient than traditional automation solutions.

How to Start Using the Agents

The new Claude agents are available on Anthropic’s financial services marketplace on GitHub. They can be used as plugins in Claude Cowork or Claude Code on all paid plans, or as Managed Agents on the Claude Platform (in public beta) for programmatic use. The new connectors and Moody’s MCP app are also available to joint customers on paid plans.

The Claude add-ins for Excel, PowerPoint, and Word are already generally available, and Claude for Outlook is coming soon.

For asset managers, banks, and fintechs that need to balance operational efficiency with regulatory compliance in a rapidly changing environment, this flexibility isn’t a nice-to-have — it’s the differentiator that determines whether the tool will still be working six months from now or will need a complete overhaul with every new regulatory requirement. The combination of specialized agents, native integration with everyday tools, and an expanding data ecosystem puts Claude in a unique position to become the standard AI infrastructure in the global financial market. 💡