How Superintelligent Interviewed 150 Staff for $500 With AI
$500
2–4 months
Not disclosed
Superintelligent needed to scale agent-readiness audits — consulting-style interviews that identify where AI agents could impact large enterprises. Traditional high-touch consulting interviews were prohibitively expensive and hard to scale, while static surveys lacked the fluidity needed for deep, natural qualitative research.
Fractional AI built a production-ready voice agent that conducts autonomous qualitative interviews at scale, asking adaptive follow-up questions guided by specific goals. It can interview hundreds of employees simultaneously with an interactive roadmap UI, live transcripts, real-time error filtering, and automatic aggregation.
A multi-agent orchestration layer handles drift detection, tool usage, and transcript filtering, using gpt-4o-realtime-preview as the primary voice agent with gpt-4o and gpt-4o-mini sub-agents and Claude 3.5 for evaluation. Custom resumption logic works around OpenAI's 15-minute session cap, and an LLM-based evaluation framework uses synthetic personas. The stack runs on React/Next.js with AWS ECS and Postgres.
Best fit for organizations that need deep qualitative interviews or discovery conducted across a large workforce faster and cheaper than human consultants.
150 interviews for $500
Thousands of interviews in month one
Built to scale to millions
Chris Taylor
Through an AI voice agent that conducts autonomous, adaptive qualitative interviews at scale. Instead of scheduling human consultants, the agent interviewed employees simultaneously with live transcripts and automatic aggregation - completing 150 interviews for one Fortune 100 client in two weeks at a total cost of $500.
The primary voice agent uses OpenAI's gpt-4o-realtime-preview, with gpt-4o and gpt-4o-mini sub-agents and Claude 3.5 for evaluation. It runs on a React/Next.js frontend with WebSockets, AWS ECS, and Postgres, plus custom logic to resume around OpenAI's 15-minute session cap.
One Fortune 100 client interviewed 150 employees in two weeks for a total of $500. The agent conducted thousands of interviews in its first month in production and is architected to handle millions of interviews per year.
Roughly two to four months from setup to production. The build included a multi-agent orchestration layer and an LLM-based evaluation framework using synthetic personas.
It is best suited to organizations that need deep qualitative interviews or discovery conducted across a large workforce faster and cheaper than human consultants.
Have a similar challenge?
