InFolk AI Leadership Assessment

Leadership assessment is broken. Most tools hand executives a static questionnaire and generate a boilerplate PDF. The insight reflects how well someone fills out forms — not how they actually lead. InFolk had a better framework — the Leadership Mirror Report, covering four dimensions: Leadership, Talent, Culture, and Strategy — but delivering it at scale required a skilled human facilitator for every session. That bottleneck capped their reach and made the product expensive to run.

The alternative was a standard chatbot, and that wasn't acceptable. Leadership conversations require nuance. A rigid Q&A flow wouldn't surface the real thinking — it would just produce answers people thought were expected. InFolk needed something that could interview like a person: following up when answers were vague, probing deeper when something interesting surfaced, and knowing when it had genuinely covered the ground.

Rebar built a conversational AI assessment platform that conducts leadership interviews through both text and real-time voice. The system uses GPT-4o to run dynamic, context-aware conversations structured around the InFolk Leadership Mirror framework. When an executive prefers to talk, they can — the app connects directly to OpenAI's Realtime API for live voice interviews with sub-second latency.

The hardest problem wasn't the AI — it was knowing when the interview was done. Long conversations drift, and a naive implementation either cuts off too early or rambles. We built a function-calling mechanism where the AI continuously self-assesses coverage across all four report dimensions before allowing submission. If a dimension is thin, it keeps going. When coverage is solid, it wraps up and hands off to report generation.

We also solved the context window problem. Leadership interviews run long. Rather than hitting token limits mid-conversation, the system tracks token budgets in real time, preserving system prompts and recent context while gracefully pruning older messages. The result is a session that never degrades — whether a session runs 25 minutes or less.

InFolk can now run leadership assessments without a human facilitator in the room. The AI delivers consistent interview quality across every session — the same probing follow-ups, the same coverage requirements, every time. What previously required scheduling a trained assessor now happens on-demand, at any hour, from anywhere.

The voice interface changed how executives engage with the process. Spoken answers are longer, less guarded, and richer than typed ones — because talking feels like a conversation, not a test. The reports reflect that. Assessors noted a marked increase in the depth and specificity of responses collected through voice compared to earlier text-only formats.