NEW YORK, 5 MAY 2026 — Anthropic's chief executive Dario Amodei sat across from JPMorgan Chase chief executive Jamie Dimon on Tuesday and unveiled something Wall Street has been quietly expecting for months: a full-stack artificial intelligence suite purpose-built for the financial services industry. The announcement, made at an invite-only briefing in New York, confirmed that Claude — Anthropic's flagship AI model — is moving decisively into the most lucrative and regulated sector in the global economy.

At the centre of the announcement were ten pre-built AI agent templates designed to automate the most time-consuming workflows in banking, asset management, and insurance. Alongside the templates came a new model — Claude Opus 4.7 — which Anthropic claims has achieved an industry-leading 64.37 per cent score on the Vals AI Finance Agent benchmark, the sector's most rigorous evaluation of AI accuracy on financial tasks. The combination of specialised tooling and an upgraded model represents the clearest signal yet that Anthropic is no longer merely a research organisation competing with OpenAI; it is a commercial enterprise targeting the estimated US$400 billion annual spend on financial services technology.

Ten Agents, One Platform

The ten agent templates span the full lifecycle of a financial professional's working day. A pitch builder assembles client presentations from internal research, earnings reports, and market data without requiring manual formatting. A meeting preparer synthesises analyst notes, news, and company filings ahead of investor calls. An earnings reviewer identifies material changes quarter-on-quarter and flags them for human sign-off. A model builder assists in constructing and updating financial models in Excel. A market researcher aggregates data from third-party sources and summarises competitive positioning. A KYC (Know Your Customer) screener cross-references identity data against sanctions lists and adverse media. A statement auditor detects anomalies across financial statements. A month-end closer coordinates the reconciliation process. A portfolio monitor tracks position-level risk and alerts on threshold breaches. A risk reviewer analyses counterparty exposure and stress-test scenarios.

What ties these agents together is their integration with Microsoft 365. Anthropic confirmed that Claude add-ins for Excel, PowerPoint, and Word are now generally available, with Outlook integration in public beta. Context generated in one application carries across to others: a financial model built in Excel can be cited directly by a PowerPoint deck being assembled in the same session, with Claude maintaining awareness of the full workflow. This cross-application memory represents a meaningful departure from siloed AI tools that require users to re-establish context with every switch between programmes.

Jonathan Pelosi, Anthropic's head of financial services, described the integrated package as designed to "close the gap" between the pace of AI advancement and financial teams' ability to actually deploy it. The candid framing acknowledges what has become a recurring frustration in enterprise AI: capability exists in abundance, but operationalisation remains bottlenecked by compliance requirements, legacy infrastructure, and institutional caution.

New Data Connectors Change the Research Landscape

Perhaps the more consequential announcement for working analysts is the suite of new data connectors Anthropic has built for Claude. The platform now integrates with Verisk, Third Bridge, Dun & Bradstreet, Experian, GLG, Guidepoint, and IBISWorld — essentially the reference data stack that institutional investors rely on daily. Rather than logging into seven separate platforms and manually synthesising results, Claude can query them simultaneously and return a synthesised brief.

The marquee partnership is with Moody's, which will embed Claude's full platform into its analytics suite covering 600 million companies worldwide. The scope of this integration is difficult to overstate. Moody's credit ratings and financial data are used by every major bank, insurer, and asset manager on the planet. By embedding Claude at this level, Anthropic gains distribution through one of the most trusted data brands in finance — while Moody's gains the ability to offer conversational analysis of its own data to clients who currently navigate it through static portals.

The market's response to the Moody's announcement was swift and stark. FactSet — whose core business is the aggregation and presentation of financial data — saw its shares fall 8.1 per cent on the day. Morningstar declined more than 3 per cent. S&P Global and Moody's itself experienced sell-offs as investors recalibrated the future of data distribution in an AI-native world. The consensus reading: if financial professionals can query a conversational AI to synthesise data from multiple sources, the value of owning a proprietary data portal diminishes.

A Joint Venture for Mid-Market Finance

Anthropic also confirmed the structure of a new joint venture targeting mid-market financial institutions — the thousands of regional banks, wealth managers, and independent broker-dealers that lack the engineering resources to configure AI systems themselves. The JV, backed by private equity firms Blackstone and Hellman & Friedman alongside Goldman Sachs, will package Claude's capabilities into a managed service requiring minimal technical deployment. The private equity backing suggests the JV is designed to be a standalone commercial entity, not merely a product division, with its own go-to-market and client success infrastructure.

FIS, the payments and financial technology giant, provided a real-world preview of what this infrastructure can deliver. In a prepared statement timed to the announcement, FIS said: "Together we're building an agent that compresses AML investigations from days to minutes." Anti-money laundering compliance is one of the most labour-intensive workflows in financial services, requiring investigators to correlate transaction records, identity documents, adverse media, and sanctions databases across multiple jurisdictions. Compressing that timeline by an order of magnitude — if the claim holds under regulatory scrutiny — would represent a material operational advantage for any financial institution.

Sector-Wide Momentum — and Realistic Expectations

The announcement came against a backdrop of accelerating AI adoption in financial services that NVIDIA captured in its 2026 Financial Services AI Survey. Sixty-one per cent of financial services firms are now using or actively assessing generative AI, 89 per cent report that AI has increased revenue while lowering costs, and 42 per cent are using or assessing agentic AI specifically. These figures represent a material acceleration from 2024, when most institutions were still running narrow proof-of-concept projects.

Dario Amodei offered one of his more candid assessments of AI's trajectory during the briefing. "The cone is even wider than I thought — we projected 10x growth only to see 80x," he said, referencing the compounding effect of capability improvements across the model stack. The remark is significant not merely as marketing; it signals that Anthropic is genuinely surprised by the pace at which Claude has found commercial utility beyond research tasks.

However, a clear-eyed reading of the benchmark numbers demands caution. A 64.37 per cent score on the Vals AI Finance Agent benchmark — described as "industry-leading" — means that Claude answers finance agent tasks correctly roughly two-thirds of the time. In a domain where a 5 per cent error in a financial model can trigger material misstatement, a 35 per cent failure rate is not a rounding error. Anthropic appears acutely aware of this tension. The human-in-the-loop design is not merely a regulatory concession; it is architecturally central to the product. "Users stay firmly in the loop — reviewing, iterating on, and approving Claude's work before it goes to a client, gets filed, or is acted on," the company stated explicitly.

This framing is essential for understanding what these agents actually are: they are not autonomous decision-makers. They are AI-assisted first drafts, compiled faster than any human analyst could manage, designed to be reviewed, challenged, and corrected by experienced professionals. The value proposition is time — not the elimination of judgement.

What This Means for Singapore's Financial Sector

For Singapore's financial institutions, the Anthropic announcement arrives at a moment of both opportunity and regulatory clarity. The Monetary Authority of Singapore published its first Artificial Intelligence in Financial Services (AIFS) framework in 2023, building on the FEAT principles established in 2018. Those frameworks emphasise four pillars — Fairness, Ethics, Accountability, and Transparency — that directly address the operational risks of AI deployment in regulated financial services.

Anthropic's human-in-the-loop design is architecturally aligned with MAS's accountability requirements. Under MAS guidance, financial institutions deploying AI must be able to explain AI-driven decisions to regulators and customers. A system that places human review before any client-facing output or regulatory filing satisfies this requirement structurally, not just aspirationally.

Singapore's three major banks — DBS, OCBC, and UOB — are among Asia's most aggressive early adopters of enterprise AI. DBS has publicly committed to over S$500 million in technology investment annually and has deployed AI across credit scoring, fraud detection, and customer service. For a bank of DBS's scale and technological maturity, Anthropic's ten agent templates represent ready-made infrastructure for workflows that currently require bespoke development. The month-end closer and statement auditor templates, in particular, address workflows that DBS's finance and compliance teams execute at significant cost every quarter.

The Moody's integration carries specific significance for ASEAN credit markets. Credit rating data on emerging market corporates in Southeast Asia has historically been less comprehensive and less accessible than data on developed market equivalents. A Claude-powered interface querying 600 million Moody's-rated companies would give regional portfolio managers dramatically improved access to emerging market research without requiring dedicated analyst headcount in each market.

Singapore's asset management sector — which managed US$4 trillion in assets under management as of 2024, making it the largest fund management hub in Asia — is a natural early adopter for the pitch builder, market researcher, and portfolio monitor templates. Firms managing capital across ASEAN's eight primary markets face precisely the multi-source data synthesis problem that Claude's integrated connectors are designed to solve.

Looking Ahead

The financial services push comes as Anthropic is reportedly preparing for an initial public offering in 2026. Financial services is not merely a vertical for the company — it is the revenue pillar that will determine the valuation at which Anthropic goes public. With Google having invested over US$2 billion in Anthropic and Amazon having committed US$4 billion, the pressure to demonstrate commercial returns in high-margin enterprise sectors is substantial.

The Blackstone and Goldman Sachs backing of the mid-market JV is a structural hedge against the risk that only the largest institutions can afford bespoke AI deployment. By creating a managed service vehicle for the mid-market, Anthropic is building distribution through existing institutional relationships — the same strategy that Salesforce used to scale into enterprise software two decades ago.

Whether Claude's 64.37 per cent benchmark score improves rapidly enough to satisfy the compliance standards of institutional finance remains the defining uncertainty. Financial services regulators move deliberately, and the reputational stakes of an AI error in a client-facing document or a regulatory filing are asymmetric. But the direction of travel — from research tool to financial infrastructure — is now unmistakably set.

Sources