Edge Shifts to Proprietary Interview Design

Diving deeper into

Joe Kim, CEO of Office Hours, on the end of crowdwork

Interview
There's a chance that an AI backed by a new generalized model will conduct interviews as well as a highly trained hedge fund analyst AI interviewer or AI agent.
Analyzed 6 sources

The durable edge in AI interviewing is moving away from the interview engine and toward proprietary question design. Voice agents are becoming cheap infrastructure, so the hard part is not getting an AI to ask follow ups, it is encoding a fund or product team’s real research taste, what they care about, how they probe, and what signals they treat as meaningful. That makes internal transcript archives and analyst workflows more valuable as training data than a generic off the shelf interviewer.

  • Office Hours already treats AI interviewing as a commodity feature, not a moat. The company says voice to AI APIs are easy to implement, experts like the asynchronous format because they can pause and resume, and the bigger effect is market expansion because more interviews can happen without scheduling a live 60 minute call.
  • Tegus shows the alternative model. Its main asset became the transcript library and the workflows wrapped around it, not the call itself. In practice, customers paid far more for searchable accumulated research than for a single interview, which suggests future value pools sit in proprietary corpuses, synthesis, and integration into analyst workflows.
  • The limiting factor is still question quality. Tegus found that calls seeded without a real thesis produced weak transcripts, because generic interviewers asked worse questions than investors with a live need. That supports the idea that a hedge fund will want an agent trained on its own analysts’ historical questions, style, and investment context, not a shared market standard.

This points to a split market. Generalized voice models will make interviewing software ubiquitous, while the highest value firms build private research agents on top of their own transcripts, playbooks, and decision processes. The winners will be the platforms that capture expert interactions cleanly and feed them back into customer specific systems of research, not the ones that rely on the interview bot alone.