Wajo AI

Wajo AI is a San Francisco startup launching Fo, a personal AI agent aimed at consumers who want errands and bookings actually completed rather than drafted. Fo comes with its own inbox, phone number, voice, and credit card, along with a builder that lets anyone create an agent of their own . The twist compared to other personal assistants is that Fo employs humans to do tasks that AI cannot , routing to vetted people in the loop whenever a step falls outside what the model can reliably handle. The launch matters now because the product is moving from closed alpha to open signup. In recent weeks Fo has ordered dinner from local delivery apps in three countries and four languages, sent a birthday gift overseas, called a cafe to check the wait for a table, booked a blowout and a haircut with a specific barber, checked a user in for a flight, compared hotels under a nightly budget, and renewed a domain registration . Wajo reports internal benchmarks where Fo completed 71% of tasks in 104 real-world tests, outperforming a rival at 42%, and passed 94% of safety checks , numbers the team argues reflect the gap between demo agents and ones that transact. Wajo is led by founder Shivani Poddar, previously Head of Engineering on Jules Agent at Google DeepMind, with earlier stints at Meta and LLM research at Carnegie Mellon , and co-authored research with team member Jeroen Olieslagers. For founders and operators watching the agent space, the launch is a concrete test of whether giving an agent real-world primitives like a phone number and a payment card, plus a human fallback, closes the reliability gap that has kept most consumer AI stuck at suggestions.