Analysis
Vinod Khosla is publicly backing the bet that Wajo, an AI agent built by former Google and DeepMind engineer Shivani Poddar, will win the crowded consumer agent market on trust rather than raw capability, according to TechCrunch. "I think the most important thing in an agent, because it's acting for you, is trust," Khosla said, arguing Wajo is "the most trustworthy because it is architected to focus on trust and safety first."
What Wajo actually does
Wajo operates across iMessage, WhatsApp, and the web, handling task management, scheduling, and what Poddar's team calls "life admin" — booking gifts, making calls on a user's behalf, hiring humans for tasks it can't do itself, and managing email. It also issues virtual credit cards so the agent can transact directly, a feature the company has filed a patent on. Since launching in late September 2026, Wajo says it has grown 10x and is live in 107 countries.
“Since launching in late September 2026, Wajo says it has grown 10x and is live in 107 countries.”
The competitive field
Wajo is entering a market already crowded with Meta's Muse and OpenAI's Dot and Instinct, all chasing the same consumer-agent use case from different angles. Wajo's team claims a specific edge there: compliance with Amazon's terms of service for agent-driven purchases, which it says Meta's Muse does not have. That is a narrow but concrete differentiator — agent shopping is only useful if the retailers an agent needs to transact with don't block it.
Why the backing matters
Neither the funding amount nor Khosla Ventures' involvement as an institutional investor has been disclosed; what's confirmed is angel backing from Jeff Dean, Google's longtime chief scientist, and Gokul Rajaram, an operator known for payments and growth roles at companies including Square and DoorDash. That combination — deep technical credibility plus payments-and-growth operating experience — is a specific bet on Wajo's two hardest problems: making the underlying model trustworthy enough to act autonomously, and making the virtual-card payment layer work at consumer scale.
What the claims don't cover yet
"Trust" as a differentiator is easy to claim and hard to verify before an agent has handled enough real transactions, calls, and payments to generate a track record of failures or breaches. A 10x growth claim over roughly two weeks since launch is also a small-base number — meaningful directionally, but not yet proof the product holds up at scale. The real test for Wajo's trust thesis will be the first publicized incident, whenever it comes, and how the company responds.