Vinod Khosla Backs Wajo's Fo as a Safer Personal AI Agent
Key Takeaways
- •Wajo's Fo is an AI "action agent" that operates with its own phone number, email address, and single-use virtual payment cards to handle bookings, calls, and emails for users.
- •At its debut, Fo completed 71% of 104 real-world tasks without assistance, compared with 42% for its closest competitor, OpenClaw.
- •Vinod Khosla, founder of Khosla Ventures, publicly endorsed Fo on September 28, 2026, citing its single-use payment cards, identity separation, and human oversight system.
- •Wajo's proprietary identity infrastructure allows Fo to interact with merchants without exposing users' real personal details, and the system never stores passwords.
- •The 94% trust rate comes from Wajo's internal benchmarks rather than independent testing, and the economics of using human assistants to finish incomplete tasks remain untested as the user base grows.
Most AI assistants can draft an email. Wajo's Fo can send it, call the restaurant, book the table, and pay with a card that stops working the moment the purchase is complete. Vinod Khosla, the founder of Khosla Ventures and a prominent OpenAI backer, is publicly vouching for the product—a pitch that rests less on how clever the agent is and more on whether users can trust it with a credit card.
Fo, an Agent With Its Own Identity
Wajo is a San Francisco startup founded by Shivani Poddar, a former engineer at Google DeepMind. The company is pitching a new category it calls “action agents.”
Fo, Wajo's flagship product, receives its own phone number, its own email address, and single-use virtual payment cards. Equipped with that kit, Fo can make bookings, place calls, and send emails on behalf of its user. When a task exceeds what the AI can handle, it can bring in human assistants to finish the work.
The Numbers Behind the Pitch
Fo entered an alpha program on August 25, 2026. After roughly a month of testing, it moved to broader public availability on September 28, 2026.
At its debut, Fo completed 71% of 104 real-world tasks without assistance. Its closest competitor, OpenClaw, sits at 42% by the same comparison.
On safety, Wajo's internal benchmarks show a 94% trust rate—a figure drawn from the company's own testing, which is worth keeping in mind when weighing it.
Wajo built proprietary identity infrastructure that allows Fo to interact with merchants without exposing the user's real personal details. The system is also designed so that passwords are never stored.
Why Khosla Is Putting His Name on It
Khosla Ventures has supported Wajo since its inception, and Khosla himself worked with the team on product concepts. He endorsed Fo publicly on September 28, 2026, the same day the agent became broadly available. His praise centered on three safety features: the single-use payment cards, the separation of user identities, and the human oversight system.
What It Means for the Agent Race
Wajo's approach attacks user hesitation directly. Single-use cards cap the damage from any single bad transaction. Masked identities keep merchants from building a profile of the user. Human oversight provides a backstop when automation fails.
A 71% completion rate is strong next to a competitor at 42%, but it also means close to three in ten tasks still require something more. Wajo's answer is to route those leftover tasks to people rather than letting them fail.
Open questions remain. The 94% trust rate comes from Wajo's internal benchmarks, and independent testing will matter as more users arrive. The economics of hiring humans to finish tasks also deserve attention: the model works well at small scale, and whether it holds up as the user base grows remains to be seen.
Source: CryptoBriefing
