Khosla Backs Wajo's AI Agent Fo to Win Market on Trust

What happened

Vinod Khosla has signalled his backing for Wajo, an artificial intelligence agent startup founded by a former DeepMind engineer, saying he expects the company to lead the emerging agent market by building on trust rather than raw capability alone. Central to that bet is Wajo's flagship agent, Fo, which is designed to go beyond typical software-only task execution: when a job requires action in the physical world or a capability it cannot perform itself, Fo can hire a human to complete the task on its behalf.

The disclosure, reported by TechCrunch, positions Wajo among a growing field of AI agent ventures competing to automate increasingly complex, multi-step work. What distinguishes Wajo's pitch, according to the reporting, is less about model performance and more about how the agent is designed to be trusted by the people and businesses that rely on it to get things done.

Why it matters

The agent market is crowded with tools that can plan, reason and execute digital tasks, but far fewer have solved for the moments when an agent hits the limits of what software alone can do. Fo's ability to delegate to a human worker is a pragmatic acknowledgement that autonomous systems still need reliable fallback mechanisms — and that users will only adopt agents at scale if they can trust the outcome, not just the intention.

For leaders evaluating AI agents, this points to a shift in how "capability" is being defined in this category. Rather than competing purely on model intelligence, vendors are starting to compete on the completeness and reliability of the end-to-end task — including knowing when to bring a human into the loop. That has implications for how organisations design workflows, assign accountability and set expectations with customers and employees about what an agent will actually guarantee to deliver.

The Renascence take

Most coverage of agentic AI focuses on autonomy — how much a system can do without human involvement. Wajo's approach quietly reframes the question: the real differentiator may be how gracefully an agent fails, or hands off, when autonomy runs out.

Trust in an AI agent is built less by what it can do alone and more by what happens at the edge of its competence. A system that can recognise its own limits and route the task to a human, seamlessly and transparently, resolves the exact anxiety that keeps most customers and employees from fully delegating to AI today. Operators building or buying agents should scrutinise this handoff moment specifically — not the demo where everything works, but the ten percent of cases where it doesn't — because that is where trust is actually won or lost.