Vinod Khosla believes ex-Deepmind engineer’s Wajo will win agent market on trust

AI-agent startup Wajo says its Fo agent is used in 107 countries and has grown 10x since launching in late September, though it did not disclose user counts or funding raised. The product offers task execution across messaging platforms, calls, virtual-card payments, and human assistance, with privacy and trust positioned as core differentiators. Early testing showed mixed performance on calls, and Wajo says it has applied for a patent covering its virtual-card system.

Analysis

The investable signal is not that a small entrant has established a trust advantage; it is that trust may become a gating condition for agents handling payments, communications, and personal data. Wajo’s human handoff and virtual-card controls could reduce user friction, but also add operating cost and limit automation economics. Its claimed growth is hard to underwrite without the starting user base, retention, task completion, and human-escalation rates; a patent application is not evidence of a defensible moat.

The more durable beneficiaries may be distributors and control points. Apple could gain if users prefer OS-level permissioning and privacy controls, while also facing pressure if third-party agents weaken app engagement. Amazon’s terms and access policies are a practical constraint on agent commerce; a stated compliance posture is not independent validation. Meta’s trust perception is a potential adoption headwind, but Khosla’s anecdotal criticism is not evidence of measurable user or ad-revenue loss. Google faces talent and product competition, though the article provides no basis to quantify impact.

Over the next 1–3 months, watch for agent access restrictions, disclosed-call standards, and evidence that users delegate consequential tasks. Over 6–18 months, economics hinge on repeat usage, safe permissions, and whether human escalation declines. The contrarian point: trust is necessary but may not be sufficient—distribution, reliability, and task coverage may matter more once baseline safeguards are accepted. Any privacy incident or unclear impersonation disclosure could rapidly reverse adoption; conversely, independently verified retention and low escalation would strengthen the startup thesis.

Key Decisions for Investors

• No directional trade on META from this article alone. Treat trust commentary as a weak sentiment signal, not a quantified threat to the ad model; revisit only if product adoption data or guidance indicates user engagement or monetization pressure.
• Keep AAPL on the relative-beneficiary watchlist, not as an immediate buy: look for evidence that OS-level controls or agent permissions increase user retention or services engagement. Falsifier: third-party agents gain broad adoption without requiring Apple-specific trust or distribution advantages.
• Monitor AMZN for policy and conversion signals rather than assuming agent blocking is bullish or bearish: tighter access could protect the customer relationship but constrain agent-driven shopping. Escalate if restrictions or agent-originated commerce materially affect seller access or reported conversion.
• For Wajo and the private-agent category, request verified active users, retention, task-completion rates, human-escalation cost, and privacy controls before underwriting growth. Broad launch claims without these metrics are not a public-equity catalyst.