Vinod Khosla bets on Wajo's AI agent, betting on trust over scale

The personal-agent space has become intensely competitive in recent weeks, with Meta’s Muse, OpenAI’s Dot and Instinct taking an early lead and new competitors launching nearly every week. But investor Vinod Khosla is betting on former Google and DeepMind engineer Shivani Poddar’s Wajo for one reason: trust.

“I think the most important thing in an agent, because it’s acting for you, is trust. You have to have trust in the agent, and I think compared to either Muse or Instinct, Wajo is the most trustworthy because it is architected to focus on trust and safety first, and some of the others are not,” Khosla told TechCrunch over a call. “Who wants to give all their data to Facebook [Meta]? I don’t. And if I give my data to Facebook, I’m the product… So trust is important, and most people don’t trust a large company like Facebook,” he added.

In broad strokes, Wajo’s “Fo” agent works much like Muse or Instinct: it operates across iMessage and WhatsApp, with a web interface organizing potential tasks into categories like “scheduling,” “gifting” and “life admin.” The agent can also call businesses or other people on a user’s behalf to book or cancel services. In early tests, the agent successfully called an airline to check for upgrades, but a call asking a user’s partner to feed the cats didn’t go smoothly, with disclosures less clear than the task required. As rivals including Instinct and Muse roll out calling abilities of their own, the industry is likely to face more scrutiny over how an AI calling on someone’s behalf should disclose itself.

One distinctive feature: Fo can hire a human to complete a task when needed. It can also create a virtual credit card to pay on a user’s behalf without disclosing personal details, and it has its own email address to converse with people and businesses. When a website or service blocks AI agents, Fo can tell the user to take action, or create a group chat with the user and the business for verification.

The startup has secured backing from Jeff Dean and Gokul Rajaram but hasn’t disclosed how much it has raised. Wajo didn’t share user numbers but said it’s being used in 107 countries and has grown 10x since its late-September launch.

Before founding Wajo, Poddar worked in engineering roles at Meta, Google and DeepMind, and was involved in building the first Facebook assistant. “The whole journey of bringing Gemini to the world and building safety for it made me realize the technological depth that you need to do this,” Poddar told TechCrunch over a call. She said trust has to be built “from day zero, not only in your architecture, but in product choices,” citing unlearning — teaching agents to forget data they’ve already accessed — as a problem the industry hasn’t solved.

Wajo has applied for a patent on its virtual-card system, and for calls, the company said it will include a “trusted contact” list for users whose agents can call them; if a call is unwanted, the recipient can tell the agent not to call again, and Wajo records that preference. The startup also said its architecture follows Amazon’s terms of service, even as tools like Muse have been blocked by Amazon. Wajo said it will explore in the coming months how Fo works across multiple users in social settings while keeping privacy intact.

Why it matters

As the AI-agent market fills with well-funded entrants from Meta and OpenAI, Khosla’s bet suggests trust and safety architecture — not feature parity — could become the deciding factor for which agent users let act on their behalf.