Level AI · 2025 - 2026
Rebuilt Voice of Customer from a manual exploration tool into an automated insight-discovery experience for leaders and analysts.
Product point of view
An AI insight product should not stop at classification. It must show what changed, why it matters, the evidence behind it, and where the user should investigate next.
01 / Context
CX leaders needed fast answers and a clear narrative, while analysts needed precise filters and drill-down. The legacy experience exposed data but made both personas work too hard to find root causes and build a defensible story.
02 / Decisions
A decision register connecting product direction to the evidence and constraints behind it.
Designed a dual-persona experience: automated narrative insight modules for leaders and deep exploration tools for analysts.
Combined a customer-editable topic hierarchy with semantic classification, dynamic themes, evidence, and embedded Ask AI.
Sequenced rollout from limited availability to beta and GA, using early customers to validate insight quality and fast-follow workflows.
Designed a self-serve setup path that generated initial themes and removed engineering from the critical path for proofs of concept.
Tradeoffs held in view
The tension stayed visible. The choice made it actionable.
Choice
Designed a dual-persona experience with narrative insight modules for leaders and deep exploration tools for analysts.
Choice
Combined generated starting themes with a customer-editable taxonomy and evidence-backed classification.
03 / System
A reconstruction of the product flow from raw conversation signals to explainable narrative insights and investigation.
Sanitized reconstruction
04 / Outcomes
Next system / 04