On May 5, Anthropic held an invite-only briefing in New York with JPMorgan CEO Jamie Dimon on stage. The agenda: a suite of purpose-built Claude AI agents, now live across Goldman Sachs, JPMorgan, Citi, AIG, Carlyle, and Visa.
The financial sector got there first. Your industry is next.
That's not a prediction — it's the explicit direction Anthropic laid out this week. And understanding what just happened in banking tells you a lot about what AI deployment actually looks like when organizations get serious about it.
This Isn't the AI You've Been Using
Most businesses that have "adopted AI" are using it the same way: open a chat window, type a question, get an answer. Useful, but nowhere close to what Anthropic just deployed on Wall Street.
The 10 new Claude agents Anthropic launched this week are a different category. Each one is pre-configured with three components:
- Skills — domain-specific instructions and knowledge for a particular task
- Connectors — governed access to the data that task runs on (credit databases, financial filings, regulatory records)
- Subagents — additional Claude models that handle specific subtasks, like comparables selection or methodology checks
The result is an AI that doesn't just answer questions — it works inside a workflow. A credit analyst doesn't describe what they need; they hand a deal to the agent and get a completed credit memo back, with sourcing, structured according to the firm's own approval process.
FIS, which handles transactions for roughly 12% of the global economy, described the difference plainly: "Every bank in the world wants AI that acts, not just assists."
The Work Looks Familiar
Here's the part worth paying attention to if you're not a banker.
The specific tasks Anthropic built these agents for — pitchbooks, research summaries, credit memos, client onboarding documents, compliance reviews — are the same types of work that exist in nearly every professional services business. The names are different. The structure is identical.
- A credit memo is a proposal with a risk framing.
- A pitchbook is a client presentation built from data.
- A KYC (know-your-customer) workflow is a structured intake and verification process.
- An earnings analysis is a competitive research report with a defined format.
If your business produces client deliverables, intake documents, research, or reports — and most do — the underlying workflow is the same one banks just got AI agents for.
JPMorgan's CEO made this concrete at the event. Over the weekend before the briefing, Dimon logged into Claude Code himself and asked it to build a dashboard covering asset swaps, Treasury bid-ask spreads, and investment-grade data. Twenty minutes later, he had a complete dashboard with sourced backup research. His reaction: "It was very accurate about what I wanted."
That's not a use case built for Wall Street. That's a use case that works anywhere someone needs to pull information together, build something structured from it, and present it clearly.
The Mid-Market Signal
The part that got less coverage: Anthropic also announced a joint venture this week with Blackstone, Hellman & Friedman, and Goldman Sachs specifically to bring Claude into mid-sized businesses.
This isn't incidental. Anthropic's strategy has two explicit tracks — one for large institutions that want to configure and run agents themselves, and one for the market below that, where the firm will embed Claude directly into company operations. Applied AI engineers from Anthropic work alongside the client, building agent workflows for the company's specific processes.
Nate Porter, a director at West Monroe, put the gap clearly: "The largest institutions have the ability to explore this space at a scale and with a degree of experimentation that a mid-market bank doesn't necessarily have the funding to be on the leading edge of."
Anthropic's answer is to close that gap. The mid-market joint venture is the mechanism. And mid-market in Anthropic's context isn't a startup with 10 employees — it's the range of businesses that have real workflows, real data, and real processes that AI can run inside of, but haven't had the resources to figure out how.
What "Agent-First" Actually Requires
The banks that are getting the most out of Claude didn't hand it to employees and say "figure it out." They redesigned the workflow first.
Goldman Sachs CIO Marco Argenti described three sequential waves at the firm: first, empowering the technical team to operate at a faster pace; second, rebuilding operational processes end-to-end; third — and in his view the most significant long-term — using AI to make better risk and investment decisions. Each wave depends on the one before it.
Most businesses are stuck between wave one and wave two. They've given employees AI tools. They haven't asked the harder question: what does this process look like if we design it around what AI can actually do?
That's a product and systems question before it's a technology question. And it's where most implementations either compound or stall.
The Practical Takeaway
The financial sector's early adoption of Claude agents gives the rest of the market a clear view of what enterprise AI deployment looks like when it's working:
- Specific tasks with defined inputs and outputs, not open-ended chat
- Real data integrations, not manual copy-paste into a prompt
- Auditable decisions with traceable steps, not black-box answers
- Workflow redesign, not tool layering on top of an old process
None of that is exclusive to banking. It's available to any business willing to get specific about what it actually needs the AI to do.
The firms that win the next two years aren't necessarily the ones that moved fastest to "use AI." They're the ones that figured out where AI fits precisely — and built the workflow around that fit.
Where SLIDEFACTORY Fits In
We've been building in the Claude ecosystem since well before this week's announcements. From SEO automation systems to agentic web platforms to client-facing tools powered by Claude, we know what the technology can do inside a real business workflow — and what it can't.
If you're trying to figure out where AI agents fit in your operations, what's worth building versus what's worth buying, or how to move from "we use ChatGPT sometimes" to something that actually changes how your team works — that's a conversation we're set up to have.
Sources
Want to figure out where Claude AI agents fit in your business operations? SLIDEFACTORY's Portland AI agency partners with mid-market businesses to design and deploy custom Claude agents, agentic workflows, and AI integrations — the same patterns the major banks just deployed, scoped for businesses your size.
Read our pillar guide on building an AI workflow stack for your business for the full architecture.