How Superintelligent Interviewed 150 Staff for $500 With AI

Superintelligent's agent-readiness audits once needed costly consultants. An AI voice agent now runs adaptive interviews across a workforce — 150 employees in two weeks for $500.

$500

Ran 150 interviews in two weeks

2–4 months

Implementation Time

Chris Taylor

CEO & Co-Founder

Fractional AI

Connect ↓

The Challenge

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.

What They Built

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.

AI Role

A primary voice agent (gpt-4o-realtime-preview) conducts each interview, adapting its follow-up questions to the respondent while staying on defined goals. Sub-agents handle drift detection, tool use, and transcript filtering, and a Claude 3.5 evaluation layer with synthetic personas scores interview quality, allowing hundreds of interviews to run at once.

AI Model

OpenAI gpt-4o-realtime-preview, gpt-4o, gpt-4o-mini, Claude 3.5 (evals), React.js, Tailwind, WebSockets, Next.js, AWS ECS, Postgres, Graphile worker

Infrastructure

React.js/Next.js frontend with Tailwind • WebSockets for live transcript streaming • AWS ECS hosting • Postgres database • Graphile worker for job processing

Integration Points

WebSocket streaming between the voice agent and the interview UI • Custom session-resumption logic around OpenAI's 15-minute Realtime cap • Multi-agent orchestration layer (drift detection, tool use, transcript filtering) • Graphile worker queue on Postgres for aggregation

Impact

150 interviews for $500

A Fortune 100 client interviewed 150 employees in two weeks at $500 total cost.

Thousands of interviews in month one

The agent conducted thousands of interviews in its first month in production.

Built to scale to millions

Architected to handle millions of interviews per year.

Technology Utilized

Implementation Complexity

Cost Reduction

Time Savings

Best Fit For

Best fit for organizations that need deep qualitative interviews or discovery conducted across a large workforce faster and cheaper than human consultants.

Chris Taylor

CEO & Co-Founder

Fractional AI

CEO & Co-Founder of Fractional AI, helping PE firms and portfolio companies implement AI workflow automations, product features, and diligence at scale.

Get an intro

Industry:

Technology & Software

Business Function:

Product & Engineering

Company Size:

Project Cost:

Cost not disclosed

Ownership:

Organization Type:

AI Pattern:

Conversational AI (Chatbot / Agent)

Natural Language Processing

Value Type:

Cost Reduction

Time Savings

AI Model:

Frequently Asked Questions

How did Superintelligent interview 150 employees in two weeks for $500?

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.

What AI models and tools power the voice interview agent?

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.

What results did Superintelligent achieve with the AI interview agent?

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.

How long did it take to build the AI voice interview agent?

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.

Who is this AI voice interview approach best for?

It is best suited to organizations that need deep qualitative interviews or discovery conducted across a large workforce faster and cheaper than human consultants.