The Money Move: Why AI Just Declared War on Finance

Act II: The Money Move

9:49

Coding was the opening act. Everyone saw it happening in real time — the benchmarks shifted, the startups multiplied, the tools proliferated, and within about 18 months the entire software engineering profession had to reckon with a permanent change to its operating model. The people who paid attention early got leverage. The people who ignored it got disrupted.

The same playbook just started running again. Same labs. Same signals. Different industry.

The target is finance.

How You Know What's Coming Next

The Frontier AI labs — OpenAI, Anthropic, Google DeepMind — don't exactly operate in silence. They telegraph their moves through benchmarks, partnerships, and product announcements before the general public realizes what's happening.

About 12 to 18 months before the coding wave crested, the benchmark cadence shifted. More evals on code generation. More research on code reasoning. More startups building developer tooling. More enterprise partnerships with engineering teams. The signal was there.

Watch the same pattern now, but pointed directly at financial services.

OpenAI has been showcasing financial reporting capabilities. Anthropic plugged Claude directly into Excel. Benchmark categories are quietly shifting toward financial reasoning, audit logic, and structured data analysis. The labs aren't making it a secret. They're calling their shots.

The Opening Salvo: OpenAI Goes Consumer

This week, OpenAI announced that ChatGPT Pro users can now connect their financial accounts through Plaid — supporting over 12,000 financial institutions. Banks, credit cards, investment accounts, retirement funds. All of it. Once connected, ChatGPT can show portfolio performance, spending patterns, subscription tracking, upcoming payments, cash flow analysis, and budget planning. It's not a chatbot with a finance plugin. It's the beginning of a personal financial operating system.

The experience, once you've had it, is apparently hard to give up.

QuickBooks lets you click around and generate reports. An AI agent lets you ask a complicated question, disappear for ten minutes, and come back to a real answer with context about your actual life. That's not a feature comparison. That's a category shift. People who've been running their finances through AI agents — exporting data, building local databases, running analysis — describe it as having a CPA-level conversation at zero incremental cost. Bookkeeping tasks that would take an accountant a few hundred dollars a month are running in under ten minutes. Tax strategy reviews that used to require a scheduled appointment now happen in a chat window. The output isn't perfect — but it's 90 to 95 percent of the way there, at a fraction of the cost, available at 2am on a Sunday. That math doesn't favor the incumbents.

The Enterprise Play: Anthropic Goes to Wall Street

While OpenAI is moving on consumers, Anthropic is going after the institutional layer — and the moves are significant.

The Blackstone joint venture. A $1.5 billion partnership with Blackstone, Hellman & Friedman, and Goldman Sachs to embed Claude inside major private equity portfolio companies. This isn't a pilot program. It's infrastructure deployment at institutional scale.

The FIS partnership. FIS runs payment infrastructure for roughly 12 percent of the global economy. Anthropic is building AI agents for financial crime investigation with them — starting with anti-money laundering. The initial accuracy rate was 64 percent. Human AML compliance accuracy sits in the mid-70s and costs roughly an order of magnitude more per decision. The question isn't whether 64 percent is good enough. The question is whether anyone seriously believes these models won't reach 75 or 85 percent in the next 12 to 18 months. The answer is obvious.

The PWC partnership. PricewaterhouseCoopers is building an Office of the CFO practice around Claude and certifying 30,000 professionals on how to use it. This is how AI becomes enterprise infrastructure — not by convincing every CFO individually, but by embedding in the consulting firms that already have those relationships. It's the same pattern that made Microsoft dominant. You don't sell to the enterprise. You sell to the firms the enterprise trusts.

At Anthropic's recent finance event — with Jamie Dimon and Dario Amodei on the same stage — a slide reportedly read: "Coding has changed forever. Finance is next."

Finance is now Anthropic's second-largest revenue category after tech. Forty percent of Anthropic's customers are financial institutions — Goldman Sachs, Visa, Citi, AIG among them. This is not a beta test. This is a land grab.

The Infrastructure Question Nobody Is Asking Loudly Enough

As AI embeds itself into banking, payments, compliance, and capital allocation, a new infrastructure layer has to be built from the ground up — one that most people haven't focused on yet.

AI agents are going to need to transact with each other. Not humans transacting through agents. Agents transacting with other agents. Fractions of a cent, moving at machine speed, across verified identities, with auditable trails. Google and Coinbase have already started building this out. Perplexity is positioning itself as the agentic Bloomberg terminal — targeting hedge funds, private equity, and wealth management with a finance-focused agent called Compute for Professional Finance.

The question of who controls the infrastructure for agent-to-agent commerce in financial services is one of the most consequential technology questions of the next decade. The answer will be determined in the next 18 to 36 months.

The Risk Nobody Wants to Talk About Directly

There's a darker side to all of this that deserves honest treatment.

Right now, simultaneously with AI's advance into financial services, we are experiencing what looks like an AI-assisted cyberattack wave unlike anything that's come before. Google reported narrowly preventing what may have been the first zero-day exploit written entirely by AI — targeting a popular open-source project with a 2FA bypass vector. Vercel experienced a supply chain breach. OpenAI disclosed that two employees were compromised. The rate of vulnerability discovery is vertical right now — more patches in the last month than in any prior comparable period.

Global financial regulators have specifically flagged AI models as potential systemic risks. The scenario they're describing isn't necessarily a direct attack. It's a confidence event. If enough financial institutions appear compromised simultaneously — or even if the fear of that reaches a tipping point — you get a panic. A bank run doesn't require actual insolvency. It requires the perception of risk. And here's the uncomfortable dynamic: the banks that are getting closest to the Frontier AI labs aren't doing it purely out of enthusiasm. There's a defensive logic embedded in those relationships. If these models are going to be the plumbing of the financial system, you want to be a customer, a partner, and a friend — not a bystander when something goes wrong.

What This Means for Microsoft (and Everyone Else)

Sir Christopher Hohn — founder of TCI, a $77 billion fund ranked among the top five in the world for compounded returns — recently sold the vast majority of his Microsoft holdings. The position was 10 percent of his fund. He cited AI as the reason. The logic isn't complicated. Microsoft's dominance was built on workflow lock-in. Excel. Word. Outlook. Teams. If AI agents replace the underlying need for those tools — if the workflow itself migrates to model-native interfaces — then Microsoft's moat is software no one needs to open anymore.

The people who think this is overstated are the same people who thought Excel would just get an AI upgrade. An AI agent doesn't use Excel better. It eliminates the reason to open it.

Dario Amodei said publicly at the JPMorgan event that some SaaS companies will go to zero. Not transform. Zero. The ones that don't adapt. That's not hyperbole from a founder trying to sell something. It's a structural prediction about what happens when the workflow layer shifts beneath the application layer.

The Bigger Frame

Here's what I keep coming back to. Coding was first because it was structured, high-value, and easy to evaluate. Finance is next for exactly the same reasons — plus regulation, which paradoxically makes it more structured, not less. The data is clean. The outcomes are measurable. The compliance requirements create documentation trails that AI can ingest and reason about.

The race has shifted. It's no longer about who has the best chatbot. It's about who controls the most important workflows. Consumer finance gives AI context about individuals. Enterprise finance gives AI lock-in on institutional workflows. Banking cybersecurity gives AI existential importance. Investment research gives AI data on how capital allocation decisions are actually made. Working with CFOs gives AI insight into how organizations prioritize and move resources.

Whoever becomes the trusted AI layer for finance gets recurring revenue, proprietary data, and deep institutional embedding simultaneously. That's the prize.

The opening act is over. We're about four acts away from the final scene and most people are still watching the curtain.

---

Rich Washburn is a technologist, AI infrastructure strategist, and founder. He advises on AI strategy, capital formation, and infrastructure deployment through Eliakim Capital and Data Power Supply.