The news: Perplexity has expanded its personal finance capabilities by supporting connections through financial technology company Plaid for checking, savings, credit card, and loan accounts for the former’s recently released Computer tool.

Zoom out: Computer looks and in many ways functions like Claude, Gemini, and ChatGPT. The distinguishing factor the company markets is the ability to break down tasks and hand them off to AI agents using optimal models to respond to each prompt. This is like asking ChatGPT to write code that pulls news from the internet and letting it handle each step in the best way possible. It is distinct from search, whose primary function is to surface, summarize, and analyze resources from the internet. The companies claim Computer can also construct personalized dashboards and create financial plans.

How it works: The broader stream of personal finance data means Computer can analyze spending, track liabilities, and calculate net worth. It builds on Portfolio, a Plaid-powered feature released in March that links brokerage accounts. This enables users to aggregate brokerage data and then query that data to analyze exposures and risk, for example.

Implications for banks: Personal financial analysis and advice have been weak spots for general-purpose AI tools, requiring manual financial data uploads. Banks and companies with direct access to consumer financial data have had the upper hand with regard to real-time data quality. But such data flowing into AI tools in real-time reduces that advantage.

Banks may have an edge because of concern about how much trust consumers will place in an assistant with access to their full financial picture. But AI adoption for financial decision-making is high—particularly among Gen Zers (77%) and millennials (72%).

The trust advantage may be short-lived: More customer activity may still migrate outside of the experience banks control. The popularity of AI bots, their growing sophistication, and—in this case—increasingly rich personal data suggests banks should either join them (i.e., make it easier to use their services with AI tools) or offer their own tools with equally sophisticated analysis and recommendations.