On May 5, 2026, Anthropic launched 10 pre-built AI agents designed to automate the most time-consuming workflows in financial services—pitchbook creation, KYC screening, earnings analysis, and financial statement audits. The agents are already deployed at JPMorgan, Goldman Sachs, Citi, AIG, and Visa, powered by Claude Opus 4.7, which scores 64.4% on financial benchmarks compared to GPT-5.5’s 59.96%. This isn’t just another AI tool release. Anthropic is making a calculated bet to become “the operating layer for Wall Street,” competing directly with OpenAI as both companies prepare for 2026 IPOs.
The market reacted immediately. FactSet shares fell 8.1%, Morningstar dropped 3%, and S&P Global saw selling pressure as investors recognized the existential threat to traditional financial data providers.
Ten Agents, Ten Workflows: From Pitchbooks to AML
Anthropic’s 10 ready-to-deploy agent templates span research, client coverage, finance operations, and compliance. Pitch Builder generates pitchbooks and comparables analyses. KYC Screener assembles compliance files. Earnings Reviewer analyzes quarterly transcripts and SEC filings. Month-End Closer manages accounting close processes. Statement Auditor reviews financials for audit readiness.
The time savings are concrete. JPMorgan CEO Jamie Dimon reported that Claude created a “comprehensive dashboard in 20 minutes with all the backup, and all the research, and it was very accurate”—work that typically takes analysts 20-40 hours. FIS is building an agent that “compresses AML investigations from days to minutes,” deploying at BMO and Amalgamated Bank in H2 2026. At Walleye Capital, a 400-person hedge fund, “100% of employees use Claude Code.”
Unlike generic AI assistants requiring custom configuration, these agents are production-ready. They deploy via Claude Cowork/Code plugins for desktop use, Claude Managed Agents for autonomous operation with audit logs, or Dispatch for voice and text task assignment from anywhere.
FactSet Drops 8%, Moody’s Pivots: The Data Provider Reckoning
If AI agents can synthesize insights directly from raw financial filings, quarterly transcripts, and public data, what’s the value of curated financial data providers? Investors answered that question on May 5: FactSet Research Systems fell 8.1%, Morningstar dropped 3%, and S&P Global and Moody’s saw sharp selling pressure.
Goldman Sachs CIO Marco Argenti captured the paradigm shift: “This represents the first time institutions can buy intelligence rather than infrastructure.” That’s a fundamental change from purchasing compute and storage to purchasing reasoning itself.
Moody’s responded with a “can’t beat them, join them” strategy, partnering with Anthropic to launch an MCP app providing credit ratings on 600M+ companies. The message is clear: traditional data providers must either integrate with AI agents or risk becoming obsolete.
64.4% vs 59.96%: Anthropic’s Benchmark Edge Over OpenAI
Claude Opus 4.7, powering Anthropic’s finance agents, scored 64.4% on the Finance Agent v1.1 benchmark—beating OpenAI’s GPT-5.5 (59.96%) and Google’s Gemini 3.1 Pro (59.72%). That 4.4 percentage point gap over GPT-5.5 shows Anthropic has a measurable edge in financial tasks.
Real-world validation backs up the benchmarks. AIG CEO Peter Zafino disclosed that “Claude scored 88% as accurate as human experts on insurance claims processing.” That sounds impressive until you consider the flip side: a 12% error rate on high-stakes financial decisions is a problem. Anthropic’s edge is real, but so are the limitations.
Vertical vs Horizontal: Anthropic’s $1.5B Bet Against OpenAI’s $4B
Anthropic and OpenAI are both racing to capture enterprise revenue ahead of planned 2026 IPOs, but their strategies diverge sharply. Anthropic launched a $1.5B joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs, betting on vertical specialization—pre-built agents for specific industries starting with finance. OpenAI countered with a $4B joint venture across 19 investors and a PwC partnership, pursuing horizontal generalization—general-purpose agents plus consulting to customize them.
Anthropic CEO Dario Amodei disclosed “80x annualized growth in one quarter” against a 10x projection, describing it as “absolute radical uncertainty” where upside scenarios continually exceed expectations. Financial services is now Anthropic’s second-largest business segment after technology. The vertical bet is paying off.
The question is whether plug-and-play vertical specialization beats customizable horizontal flexibility. Wall Street’s adoption patterns will determine the answer—and both companies’ IPO valuations.
Claude for Excel, PowerPoint, Word: Meeting Analysts Where They Live
Anthropic’s distribution strategy is smart: Claude now operates natively within Microsoft Excel, PowerPoint, Word, and Outlook via add-ins. Context automatically transfers between applications—financial models built in Excel flow seamlessly to PowerPoint presentations without re-explanation. Outlook functions as a “chief of staff” handling inbox triage and meeting scheduling.
Citadel’s Head of Core Engineering explained the value: “Claude for Excel meets investment professionals where they live—in data and analytical models. Analysts use it to build coverage models, separate signal from noise, and pressure-test work with step-change efficiency.” Financial analysts spend their days in Excel. Anthropic is embedding Claude directly into that workflow rather than forcing users to switch tools.
Key Takeaways
- Anthropic deployed 10 finance-specific AI agents at JPMorgan, Goldman Sachs, Citi, AIG, and Visa—pitchbooks in 20 minutes vs 20-40 hours manually, AML investigations compressed from days to minutes.
- Claude Opus 4.7 scored 64.4% on Finance Agent benchmarks, beating GPT-5.5 (59.96%) and Gemini 3.1 Pro (59.72%), with 88% accuracy on insurance claims processing.
- Traditional data providers took an immediate hit—FactSet fell 8.1%, Morningstar dropped 3% on May 5 as investors priced in the threat to curated financial data business models.
- Anthropic’s $1.5B vertical specialization bet (pre-built finance agents) competes with OpenAI’s $4B horizontal approach (general agents + PwC consulting), with both companies targeting 2026 IPOs.
- Microsoft 365 integration (Excel, PowerPoint, Word, Outlook) puts Claude where financial professionals work, with seamless context transfer across applications.
Both Anthropic and OpenAI are betting billions that enterprise AI adoption will define their IPO valuations. Wall Street’s choice between vertical specialists and horizontal generalists will determine which strategy wins.
