Wajo Bets on Trust for Its Agents

The startup founded by Shivani Poddar is advancing a personal agent, Fo, built around mechanisms of trust and security, from virtual cards to call control. It claims an architecture compliant with Amazon's rules and is attracting significant support. Vinod Khosla says he is betting on Wajo against Muse, Dot, and Instinct, precisely because of this trust-centered approach.

Amazon Compliance, Patent, and Announced Call Control

Wajo asserts that its architecture complies with Amazon's terms of use, while tools like Muse are reportedly blocked on the platform. The company has filed a patent application for a virtual card system designed to mask user information from websites. For calls, it plans to introduce a list of "trusted contacts" authorized for each user. A recipient can request not to be called anymore, and this preference is recorded by Wajo. The Fo agent also has a complaint solution for agents on Amazon. The company indicates it is working on several fronts around trust and security and aims to test Fo's functionality in social contexts with multiple users in the coming months, keeping confidentiality as a guiding principle. The stated challenge is to avoid privacy concerns that have affected other agents, even as security is established as a core value.

What Fo Can Already Do on Mobile, Web, and Email

Fo operates on iMessage and WhatsApp, and offers a web interface with categorized task cards, including planning, gifts, and life administration. The agent can call businesses or individuals on behalf of the user to book or cancel services. It can issue a virtual credit card for payments without exposing sensitive information. Fo has its own email address to communicate with human contacts. If a service blocks AI agents, Fo can either alert the user to take over or create a group with them and the business for verification. If necessary, the agent can also delegate task execution to a human.

Declared Traction and Support, but Few Numbers

The startup claims support from Jeff Dean and Gokul Rajaram. It does not disclose the amount of its fundraising or the number of users. However, Wajo states that its service is present in 107 countries and that its growth has multiplied tenfold since its launch at the end of September. The company does not provide further details on these figures.

A Trust-Centered Thesis, Supported by Khosla and Poddar

Vinod Khosla states he is betting on Wajo because trust is paramount for an agent acting on behalf of the user. He considers Wajo more trustworthy than Muse or Instinct, believing it is primarily designed around security, and asserts he does not want to entrust his data to Meta, a company he deems not credible for data protection. Shivani Poddar presents trust as the founding principle of Wajo and believes that agents do not yet know how to "unlearn," hence the necessity to integrate trust from day one in the architecture and product choices. She aims to design useful products, after an experience with Gemini that showed her the technical scope of the issues, and considers that Google can sometimes be behind in product development. As the founder of Wajo, Poddar has worked at Meta, Google, and DeepMind, and participated in the first Facebook assistant.

Call Function in Development and Market Under Pressure

The market for personal agents has intensified in recent weeks, with Meta's Muse, OpenAI's Dot, and Instinct among the references, while new entrants are multiplying. Fo is presented as operating broadly on the same model. Preliminary trials report a successful call to an airline to check for upgrades, but another call to a relative went poorly, with disclosures deemed less clear. Competing players like Instinct and Muse are also gradually deploying calling capabilities, which could fuel increased discussions about how an AI should present itself when calling on behalf of a user.