TL;DR
- Anthropic released 10 finance-focused AI agents on May 5th, built into Microsoft 365 tools like Excel and PowerPoint.
- They split neatly into front-office work (pitch decks, earnings review, financial modeling) and back-office work (ledger reconciliation, auditing, KYC screening).
- The interesting part isn't the agents themselves... it's that you can build similar agents with a cheaper model and the right prompts, and get 90% of the way there.
- Small businesses can adapt most of these for their own operations, from reconciling accounts to screening new customers.
- Anthropic owned the software industry, now they're targeting finance. Healthcare is probably next.
After two years of making every software developer's life easier (or more anxious, depending on who you ask), Anthropic is pointing Claude at the next big target: finance.
On May 5th, they released 10 financial AI agents built on Claude, shipping natively inside Microsoft 365. The pitch is straightforward: give analysts a junior assistant that never sleeps, never complains about month-end close, and can process an earnings report faster than any human ever could.
But the more interesting story isn't the agents themselves. It's the pattern they reveal... and what that means for businesses that have nothing to do with Wall Street.
The Front Office: Research and Client Coverage
The first five agents handle the visible, client-facing work that analysts and bankers spend most of their time on.
The Pitch Builder drafts a client deck and pulls comparable company financials automatically. The Meeting Preparer compiles the books before a call so the banker isn't speed-reading all night. The Earnings Reviewer kicks off the moment a company drops its quarterly results, pulling out the changes and writing a summary before a human even opens the PDF.
That last one is the most interesting to think about. A human analyst was never going to beat algorithmic trading systems on earnings day anyway. But now you have AI agents reading the same report and potentially trading on it in seconds. That creates a legitimate question about whether volatility around earnings season gets worse as every firm's agent interprets the same data slightly differently and acts on it immediately.
The Model Builder auto-generates discounted cash flow models and other financial projections. And the Market Researcher scans news and trends for a given sector.
Here's what I find interesting about the model builder specifically. People who've built financial models by hand will tell you the real value wasn't the finished spreadsheet. It was the hours spent inside a company's financial statements, learning where they make money, where they spend it, and what looks different this quarter. Automating that output might save time. It might also produce analysts who understand the numbers on the surface but miss the story underneath... similar to what we're seeing with developers who can ship code fast but don't understand the architecture well enough to know when it starts getting fragile.
The Back Office: Accounting, Auditing, and Compliance
The next five agents go after the slow, repetitive tasks that keep accounting teams busy every month.
The Valuation Reviewer checks existing company valuations against methodology and external data. If a company looks significantly undervalued compared to its peers, it flags it. (This is the one I'd personally want to build my own version of just to poke around the market for opportunities.)
The General Ledger Reconciler and Month End Closer are exactly what they sound like... reconciling accounts, running net asset value calculations, preparing journal entries, and closing the books. Not glamorous, but genuinely painful for anyone who does it manually every month.
The Statement Auditor is the one that actually sounds like something from a thriller. It reviews financial statements looking for discrepancies and anomalies. A company that's never had revenue from a certain category suddenly does? Flagged immediately. Whether that's fraud detection or just an unusual one-time event, the agent at least gets the question in front of a human fast.
And the KYC Screener (Know Your Customer) automates background checks and risk screening for new clients. Before a loan goes out, before a new account gets opened... run the AI, check the signals.
It Lives in the Tools You Already Use
Here's the part that makes this more than just a press release: Anthropic built these directly into Microsoft 365. The Model Builder shows up in the Excel ribbon. The Pitch Builder lives inside PowerPoint.
That's a different move than releasing a standalone API product and hoping developers build on top of it. It's meeting people inside the software they already open every morning. The barrier to adoption drops dramatically when the tool is just... already there.
If you want to start exploring what AI tools are actually worth integrating (not just the ones that make headlines), the AI tools checklist at Infacto is a good place to get oriented before you start clicking things in the ribbon.
You Don't Need Anthropic's Version
This is the mindset shift worth sitting with.
Every one of these agents is doing something you could approximate with a cheaper model, the right context, and some clear prompts. You're not going to match Anthropic's fine-tuning on financial data. But 90% of the value? Probably reachable.
The pitch builder is just "here's the company data, write me a deck." The meeting preparer is "here are the docs, summarize what I need to know." The statement auditor is "here are the financials from the last 12 months, flag anything that looks unusual compared to prior periods."
None of those are magic. What Anthropic is selling is the polish, the integration, and the trust that comes from a known brand in a regulated industry. For a big bank, that matters. For a small business trying to do its own version of ledger reconciliation, maybe it doesn't have to.
What Small Businesses Actually Get From This
These agents were built for Goldman Sachs, not your landscaping company. But the underlying logic applies everywhere.
The General Ledger Reconciler is just... bookkeeping. If you're a small business owner reconciling accounts manually, a version of this agent would handle that for you. The Statement Auditor could review statements from your suppliers and flag anything that looks off. The KYC Screener could run background checks on potential customers before you extend credit, or on candidates before you hire.
Think about what each of these agents is actually doing, then ask whether there's a version of that problem in your own business. Most of the time, there is.
If you're not sure where your business has the biggest gap between what you're doing manually and what could be automated, the small business strategy diagnosis quiz will help you figure out where to start.
Anthropic's Next Target Is Probably Healthcare
Anthropic spent the last two years going deep on software engineering. Claude Code, Claude's native integration with developer tools, the whole arc of their product roadmap has been aimed at that audience. Now finance.
The pattern makes sense. Pick a high-value, information-dense industry with lots of repetitive analytical work. Build agents that slot into the existing workflow. Ship it inside tools people already use.
Healthcare fits that profile almost perfectly. Hospitals, insurance, clinical documentation... the same combination of slow manual processes, high stakes, and enormous paper trails. My prediction is that's the next industry Anthropic makes a serious move on.
The Job Gets Bigger, Not Smaller
The real shift this creates isn't that finance jobs disappear. It's that the scope of one person's job expands.
An analyst who used to spend all their time reviewing earnings can now also build the models, research the sector, and prep the client meeting. One person gets to own more of the work because they're not buried in a single lane anymore.
That's the same thing that happened in software. Senior developers didn't disappear when AI code generation got good. They just stopped writing boilerplate and started solving harder problems. Finance is probably about to go through the same transition.
The businesses that figure out how to expand the job, not just speed up the old one, are the ones that pull ahead.