Anthropic Is Trying to Ground Rocket Money

On November 3 I published I Fired Rocket Money. Your Company Could Be Next. It was out of date in two days.

The short version of that article: I canceled my Rocket Money subscription and replaced it with a budgeting system I built myself. Rocket Money sold me a basic set of features on top of a connection to my banks. Once I could get that connection from Plaid, the company that supplies it to apps like Rocket Money, I could build the rest. My argument was that a lot of companies sitting between us and what we need are going to get cut out the same way. I also wrote this: "The good news for a lot of companies is that this won't happen fast."

Anthropic Launched Personal Finance on November 5. OpenAI Did It in May.

Two days after my article went up, Anthropic, the company that makes Claude, announced Personal Finance in Claude. You connect your bank and card accounts through Plaid, and Claude sorts your spending, flags odd charges and answers questions about your money. It's in beta, an early test version, for paying customers in the US. It can read your accounts but can't move money.

When I found it, I thought the timing was wild. Then I started researching this article and learned that OpenAI put the same kind of feature into ChatGPT back in May. I'd completely slept on it. I'm sure Rocket Money knew about it long before I wrote a word.

I Was Too Flippant About Rocket Money

So the competition I wrote about is already here, and it's coming from OpenAI and Anthropic. My first article had families like mine slowly building their own budgeting systems.

I wrote that article as if Rocket Money had plenty of time to adjust. If I'd known Anthropic was about to step up to Rocket Money, with OpenAI already there, I'd have picked a different example. Rocket Money's head of product and AI left a gracious comment on my post and offered to show me what they're building. I told him I'm rooting for Rocket. The pressure on Rocket Money is starting sooner than I said it would.

Anthropic Is Competing With Its Customers

Rocket Money is an Anthropic customer. Its AI agent, Rowan, an assistant that cancels subscriptions and negotiates bills for you by text message, is built on Claude.

Anthropic advertises that. It has a customer story on its own website about how Rocket Money built Rowan, with numbers on how much faster the team works and a quote from the same Rocket Money executive who commented on my post: "Our bar for accuracy needs to be 100%." That page is still up. So is the page announcing Anthropic's own product for tracking your money.

Rocket Money builds its product on Anthropic's AI, and Anthropic shows Rocket Money off as a success. Then Anthropic releases its own product that does part of what Rocket Money does. It has happened with coding tools too. Claude Code, Anthropic's own coding tool, competes with Cursor, a coding app that relies on Claude.

If your company builds its product on Anthropic's or OpenAI's AI, your supplier can become your competitor.

Are Anthropic and OpenAI Product Companies Now?

I don't think of budgeting tools as their business or their expertise, and I can't tell from either launch where the product is headed. Anthropic already has versions of Claude for legal and for financial services. Who's next? Adding one product after another reminds me of AOL in 1996, trying to put everything on the internet inside one app.

The products are also weak. I connected the same bank accounts to Claude and to ChatGPT and checked both against my own system. Claude counted a mortgage payment and a car payment as income. ChatGPT's spending total came up short, and its grocery number was less than half of mine. Claude told me a phone insurance claim and an iPad repair were yearly charges. Both still think I'm paying for Rocket Money.

I can see why. In my tests each one worked only from what the bank sends, and bank labels are often garbled. Both are read-only, so neither can take action for you. Rocket Money can at least cancel a subscription. My system reads my email, my texts and my location history, and one search of my email showed that the phone charge was an insurance claim. My system isn't perfect either. It missed a membership renewal that both chatbots caught. My first draft of this paragraph said I could fix that. I've already fixed it, so this article was out of date before I published it too. A custom system can be modified or fixed almost immediately.

The weakness confuses me, because my system runs on Claude. Anthropic's AI, once it could search my email, got those two charges right. Anthropic's own product, built on that same AI, got them wrong. Don't these companies see that their own AI is making one-size-fits-all software easy to replace?

I Got the Speed Wrong

I still think the companies in the middle will get cut out, unless they restructure their products to protect their moat, the advantage a competitor can't easily copy. They'd need to open their services to their customers' AI agents and focus on their unique value, not their interfaces.

The competing products are still very rudimentary, so they're not a threat today. But the AI industry keeps changing at a rapid pace, and it clearly isn't slowing down. Hold on to the steering wheel with both hands.

P.S. The Sales Pitch

Is your company at risk of AI cutting you out? If you sit between your customers and what they need, or you build your product on someone else's AI, I can help you find where you're exposed, plan what to do about it and speed up your product work with AI agents. Find me at wespomeroy.com. And if you want a system like mine, built around your own life, I build those too. Company or person, I'm here for it.