Anthropic Unveils 10 Finance AI Agents as Amodei Reverses Jobs Warning

The AI company Anthropic announced a suite of ten specialized agent templates designed to automate routine financial services work, from building pitchbooks to reconciling general ledgers. The release marks a significant expansion of the company's enterprise offerings, with financial services now representing its second-largest industry by revenue after technology.

Each agent ships as a plugin in Claude Cowork and Claude Code, alongside cookbooks for Claude Managed Agents. The templates package three components: skills (instructions and domain knowledge), connectors (governed access to data), and subagents (additional models called for specific sub-tasks). Firms can adapt them to their own modeling conventions and approval flows.

The full roster includes a pitch builder for client meetings, a meeting preparer for briefs, an earnings reviewer for transcripts, a model builder for financial models, and a market researcher for sector tracking. On the operations side, there's a valuation reviewer, general ledger reconciler, month-end closer, statement auditor, and KYC screener. Users stay in the loop, reviewing and approving Claude's work before it reaches clients or gets filed.

According to the official announcement, these agents pair with Claude Opus 4.7, which leads the industry on Vals AI's Finance Agent benchmark at 64.37%. The company also expanded Microsoft 365 integration, allowing Claude to work directly across Excel, PowerPoint, Word, and Outlook with context carrying automatically between applications.

What's more notable than the technical specs is the narrative shift from CEO Dario Amodei. Last year, Amodei was among Silicon Valley's most prominent doomsayers on AI employment, publicly warning that AI could eliminate half of entry-level white-collar knowledge work within years. That stark projection made him the rare tech founder willing to say out loud what many peers only whispered.

At the Lower Manhattan briefing on Tuesday, sitting onstage alongside JPMorgan Chase CEO Jamie Dimon for the first time, Amodei reached for a different intellectual framework: the Jevons Paradox. "If you automate 90% of the job, then everyone does the 10% of the job. And the 10% kind of expands to be 100% of what people do and kind of 10xs their productivity."

It's a more comfortable theory. It's also one he immediately complicated (the same breath that invoked optimism also acknowledged AI is moving faster than past technologies).

William Stanley Jevons was a 19th-century British economist who observed that as steam engines became more efficient and coal cheaper to use, total coal consumption went up, not down. Applied to AI and labor, the logic runs: if AI makes a lawyer 10 times more productive, legal services become cheaper; cheaper services mean more people use them; more demand means more lawyers, not fewer.

Dimon made the same argument in blunter terms, invoking agriculture, electricity, and the internet. "The capitalist society is very good at recreating jobs and recreating things," he said. "And life is better. Not always if that town loses a factory, but in general better."

The problem is that Amodei, almost in the same breath, described precisely the condition under which Jevons stops working. "AI is moving faster than all these previous technologies," he said. "And so when you strain a system more than it's usually strained, it's possible you get these weird behaviors and this big disruption."

This is not a minor qualification. The Jevons mechanism depends on time—time for markets to recognize new demand, for workers to retrain, and for employers to expand rather than simply contract. The ATM is the classic cautionary example: it didn't eliminate bank tellers immediately, but over two decades, teller employment fell sharply as branch activity shifted. AI is not operating on a two-decade timeline.

Even the optimists acknowledge that Jevons operates at the aggregate level, not the individual one. If AI expands demand for legal services globally, that's good for BigLaw partners and bad for first-year associates, whose document-review work no longer exists. The pie gets bigger; the slices don't redistribute automatically.

Amodei gestured at this problem but didn't resolve it. "Companies have a choice," he said. "They can do the same thing with less resources—and that leads to things like layoffs—or they can do more with the same amount of resources. But that requires creativity." He and Dimon both endorsed some form of wage-reassurance programs and government-funded retraining.

Dimon pointed to trade adjustment assistance after NAFTA as a model—before acknowledging that it was a pretty bad example. "It didn't work because it wasn't set up right, because it made it too hard to get the benefits. So it's solvable, but only with collaboration in government and business."

Business Insider reported that 40% of Anthropic's top 50 customers are in finance, and the company is enmeshed in an already competitive space. Startups like Rogo, valued at $2 billion and founded by former investment bankers, serve more than 250 clients with similar tools. Major banks including Goldman Sachs and Morgan Stanley have rolled out internal AI assistants to large swaths of their workforces.

The physical reality of using these agents matters. An analyst who's started a model in Excel doesn't need to re-explain it when that work moves to PowerPoint. In Outlook, Claude can act as a chief of staff that triages your inbox, arranges meetings, and drafts responses in your voice. In Excel, it builds financial models from filings and data feeds, audits formulas across linked workbooks, and runs sensitivity analyses.

Amodei's evolution on this question is worth tracking closely. When the CEO of the company building the technology starts invoking optimistic economic theory, there are two possible explanations. Either he has genuinely updated his view based on new evidence, or the social and political cost of the bloodbath framing—particularly as Anthropic navigates a Pentagon lawsuit and a fraught regulatory environment—has made it more useful to suddenly sound a bit more optimistic.

Whether users actually pay for these agents and whether the Jevons effect materializes fast enough to matter for displaced workers remains the real question.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn