Why an AI Budgeting App Doesn't Fix a $37 Billion Debt Problem

A mortgage company with $37 billion in debt just announced an AI assistant that texts you to cancel subscriptions. The headline says "rewrites what AI can do in personal finance." The balance sheet says something else entirely.

Rocket Money, a unit of Rocket Companies (NYSE: RKT), launched Rowan on Tuesday. Built with Anthropic, it watches your accounts, renegotiates bills, cancels subscriptions, and sets up savings — all through text messages. You reply "cancel" and it calls the service provider for you. No dashboard to open. No app to navigate. Just a text thread.

The product idea is straightforward. Most personal-finance tools — Mint, Copilot Money, even ChatGPT's new finance feature — show you your money and let you figure out what to do. Rowan does the doing. That's the claimed difference between knowing and acting.

It sounds useful. The question is whether it matters for the stock.

Rocket Money has roughly 5 million users according to its app store listing. A free plan tracks subscriptions and spending. The paid tier, called Premium, costs $7 to $14 a month under a pay-what-you-want model. On top of that, bill negotiation charges 35% to 60% of the first year's savings. Rowan is available through a new Premium Plus tier, with broader rollout planned later this year.

Now put those numbers next to the parent company.

Rocket Companies had $10.2 billion in trailing revenue and a $40.5 billion market cap. It carries $37.4 billion in total debt, $3.1 billion in cash, and reported $634 million in negative operating cash flow over the last twelve months. The stock is down 26% year-to-date. Rocket Money itself is not reported as a separate segment. It lives inside the "All Other" segment which includes Redfin, Rocket Loans, and Mr. Cooper, and generated $1.82 billion last quarter. That's about 6.5% of total revenue, and Rocket Money is only a piece of that.

The AI assistant is a rounding error relative to the balance sheet.

So why launch it? And what should an investor actually take from a press release that describes text-message subscription cancellation as rewriting the future of finance?

The answer is not about the product. It's about the company trying to change the label on itself.

Rocket Companies has spent years building a narrative. The pitch: we're not a mortgage originator subject to interest rate cycles. We're a fintech platform. We acquired Truebill (renamed Rocket Money) for $1.275 billion in 2021. We acquired Redfin and Mr. Cooper in 2025. We invested over $5 billion in technology. We have 25 million accounts across our ecosystem. Each acquisition and product launch is another brick in the wall between "mortgage company" and "AI-powered financial platform."

The market has not bought it. The stock trades at 86 times trailing earnings, which sounds like a tech multiple — until you notice forward earnings are negative, return on invested capital is negative 0.9%, and operating cash flow is deeply in the red. The multiples are high because the earnings base is small, not because the market is pricing in superlinear growth.

Rowan is the latest brick.

There's nothing wrong with building a product people want. If Rowan actually gets people to pay for Premium Plus — if the text-message interface is genuinely better than opening an app — that's real progress. The competitive space is already crowded. ChatGPT added personal-finance features in May, letting users connect accounts through Plaid. Meanwhile, Copilot Money charges $13 a month. Cleo, Monarch Money, and YNAB are also in the mix. They all show dashboards. Rowan's claim to do things instead of just showing things is the differentiator.

But the differentiator matters to Rocket Money's business, not to Rocket Companies' balance sheet. Even if Rowan doubles Rocket Money's revenue, it changes very little about a $10 billion company with $37 billion in debt. The AI assistant does not refinance a mortgage portfolio. It does not reduce interest expense. It does not fix negative operating cash flow.

The more useful question is whether the narrative strategy is working in reverse. A company that announces an AI agent for personal finance on the same day its stock is 26% off its highs is asking investors to believe two things at once: that this is a platform company worth a premium, and that its core mortgage business is stabilizing after the worst rate environment in decades. The financial data supports neither conclusion with any conviction.

Rocket Companies did report improving profitability over the last five quarters. Q2 2026 adjusted EBITDA hit $766 million, a 28% margin, up from 26% in Q1. Revenue nearly doubled year-over-year to $2.8 billion, though a large part of that came from acquisitions rather than organic growth. The Q2 revenue miss against consensus of $2.81 billion sent the stock lower before it recovered the next day.

What the numbers tell you is that the mortgage business is recovering from a brutal cycle — rate cuts are bringing locked and closed loan volume back up — but the recovery is thin and the leverage is enormous. Return on equity of 3% on a 1.7x price-to-book is not the profile of a company the market is excited about. It's the profile of a company the market is tolerating, waiting to see whether the debt gets paid down or the earnings base gets bigger.

Rowan does not change that waiting game.

The real test for Rocket Money is whether users actually choose a text thread over an app. The industry has been pushing dashboards for years. People link accounts, check their budget, and then forget about it. A conversational interface could work if the AI is actually useful in real time — if it texts you at 7 PM to say your trial is expiring tomorrow, or catches a subscription you forgot. It could fail if people find it annoying, or if the text interface turns out to be a gimmick that doesn't scale past the most enthusiastic early adopters.

Nobody knows the answer yet. Rowan launched today to a "select subscriber" group. The broader rollout is planned but undated. There are no adoption numbers, no retention data, no revenue attribution. The claims come from company executives describing capabilities, not from users describing behavior.

For the RKT shareholder, the right frame is simpler than the press release suggests. The AI assistant is real. The product may or may not be good. Neither conclusion changes the fact that you own a mortgage company with a mountain of debt, and the mortgage business is what will determine whether this stock goes up or down over the next year. The fintech narrative is interesting. The balance sheet is what you're actually exposed to.

Watch whether Rowan actually moves the needle on Rocket Money's subscription revenue over the next two quarters. Watch whether the "All Other" segment grows fast enough to be reported separately. And watch the mortgage side — lock volume, EBITDA margins, debt paydown — because that is the business that matters for the price of the stock.