Fi Money rolls out AI feature for users' queries on personal finance via ChatGPT, Gemini

Peak XV and Temasek-backed fintech platform Fi Money has launched a new feature that allows users to securely connect their financial data to AI assistants such as ChatGPT, Gemini, and Claude.

The integration enables users to ask personalised questions about their money from spending patterns to investment performance and receive real-time, AI-generated insights based on actual financial data. The new feature, called Talk to AI, is powered by Fi’s implementation of the Model Context Protocol (MCP), a framework that acts as a secure bridge between a user’s consolidated finances and external AI tools.

With this, Fi users can export or stream their financial information such as bank accounts, mutual funds, loans, insurance, EPF, and stock holdings  to supported AI models.

“You can ask AI any personalised question about your financial data,” said Sumit Gwalani, co-founder of Fi Money to Moneycontrol. “People are calling it their CFO or their CA.”

How does it work? 

Until now, users typically interacted with generative AI tools such as ChatGPT by entering generic prompts or manually typing in data. Fi’s new infrastructure claims to follow a structured, consent-based method of sharing financial data, reducing the need for guesswork and improving the quality of responses.

Users can either share an AI-optimised export file through the app or set up a "secure tunnel" using Fi’s MCP server, a developer-friendly interface that streams near real-time data to AI tools.

“We’re building a secure tunnel between your financial data and the AI assistant,” Gwalani said. “Normal ChatGPT works, you don’t need a Pro version.”

Fi said privacy is built into the system. Personally identifiable details such as PAN and voter ID numbers are sanitised before data is shared with the chatbots. Two-factor authentication is also required when using the MCP server.

The company has been testing the tool with early users, reporting improved accuracy and lower hallucination rates.

“In our tests, hallucinations dropped by 90–95 percent because we clean and structure the data before it reaches the AI,” Gwalani said. “We also pre-calculate some values to help the AI generate more precise responses.”

While the tool offers read-only functionality, Fi isn’t ruling out enabling transactions or money movements in the future.

“We could let users say, ‘Move Rs 10,000 into this mutual fund,’ but there are regulatory questions. We’re cautious about crossing that line without guidance,” Gwalani said.

The company also sees the AI integration as a step toward broadening its user base beyond core banking customers. Fi originally required users to open a savings account to use the app, which is no longer mandatory.

“Earlier you had to open a a Federal bank account on the Fi app. Now, you can use Fi just to track your finances or interact with AI, without needing to open an account,” Gwalani said.

AI-enabled fintech services on a rise 

The launch comes amid rising interest in AI-powered finance in India. Zerodha’s Rainmatter recently introduced a similar MCP-based integration that allows users to connect investment data to Claude and Cursor.

Also Read: Zerodha’s AI foray wins early fans, but raises questions over assisted investing

Fi tracks over Rs 10,000 crore in assets. The company serves more than 3.5 million users across 19,000 pin codes.

Founded in 2019 by former Google Pay (Tez) executives Sujith Narayanan and Sumit Gwalani, Fi offers savings accounts, instant loans, mutual funds, credit cards and US stock investments in partnership with regulated financial institutions.

The MCP integration is now available to all Fi users via the app. Developers can also set it up independently using third-party AI tools like Claude or Cursor.

Financial data security

Gwalani said the system has been designed with user privacy and data security at its core. “We sanitise PAN numbers, voter IDs, and other personal identifiers before data goes to AI,” Gwalani said. Users must also complete two-factor authentication when setting up the MCP server and are re-authenticated periodically to prevent persistent access.

“We’re designing this in a privacy-preserving way,” he added. Fi plans to roll out controls that will allow users to choose which parts of their financial data to share with AI and which to withhold.