This month OpenAI acquired Hiro, an AI financial planning app that billed itself as an "AI personal CFO" and helped clients manage more than $1 billion in assets[1]. Terms were not disclosed, the deal was characterised as an acqui-hire, and Hiro stopped operating on 20 April[2]. It follows OpenAI's purchase of another personal finance app, Roi, in October 2025.

On its own, a small acqui-hire is not news for banks. The direction it points in is.

The shift: advice before accounts

Hiro's founder, Ethan Bloch, who previously founded the savings app Digit, put the ambition plainly: "For decades, personalized financial guidance has been too expensive, too generic, or too hard to access. ChatGPT is finally changing that"[1].

Banks have always assumed that owning the account means owning the relationship. That assumption held because the account was where customers saw their money and made decisions. A general-purpose AI assistant changes that. If the natural place to ask "can I afford this holiday?", "should I overpay my mortgage?" or "why am I always short at the end of the month?" becomes a chat window, the first conversation about a customer's money happens outside the bank.

Whoever answers the question "can I afford this?" owns the relationship. The account behind the answer becomes a utility.

What the assistant still needs from banks

The assistant's advice is only as good as its data. Planning tools like Hiro rely on users typing in or connecting their salary, debts and outgoings. That puts banks in a strong position, for now: they hold the transaction data, they grant or refuse data access, and they execute the payments and transfers that advice leads to.

The question is how long that leverage lasts. Open banking rules, customer-permissioned data sharing and agent-initiated payments all make it easier for an assistant to see the account and act on it without the customer ever opening the bank's app.

Three moves for banks

  1. Make the banking app the best place to ask a money question. Customers already trust the bank with their data. An in-app assistant that answers with the customer's real numbers, and can act on the answer, is a better experience than a general assistant working from estimates.
  2. Decide your data-sharing stance deliberately. Refusing access protects the relationship briefly but frustrates customers. Allowing it on clear terms, with visible permissions and reliable connections, keeps the bank as the trusted source.
  3. Design products that assistants can recommend. If advice increasingly comes from AI, products need clear, machine-readable terms and pricing. An assistant cannot recommend a savings account whose rate it cannot find.

The strategic question

Personal finance has been one of the most discussed use cases for generative AI since the start. The acquisitions suggest the largest AI companies intend to build it properly. Banks should decide now whether they want to be the adviser, the trusted data source for someone else's adviser, or simply the account the adviser moves money out of.