Our AI platform required evolution toward generative capabilities, but risked losing major enterprise clients (Verizon, JP Morgan) who depended on our data labeling disambiguation workflows. The challenge: maintain familiar interaction patterns while enhancing functionality in a cleaner, more focused interface.

✓ Preserved enterprise relationships with seamless transition that maintained all client contracts

✓ Successful client migration from legacy to new platform with zero workflow disruption

✓ Product Owner feedback: "It went really smooth and worked perfectly"

✓ The disambiguation feature redesign became the final piece enabling our biggest clients to adopt the new platform

The redesign transformed a cluttered interface with scattered similarity metrics into a purpose-built disambiguation tool that:

Clarified context - Users now immediately understand they're in disambiguation mode with clear indication of what intent they're working with

Improved information hierarchy - Separated conflicting intents from utterances with clear visual distinction

Enhanced data representation - Redesigned conflict percentages to be more scannable with visual indicators

Simplified decision-making - Created a cleaner layout that focuses attention on resolving conflicts

Conducting deep user research with power users and employing an atomic research approach was fundamental. I designed a solution that preserved critical NLU capabilities while seamlessly integrating new generative AI features - including contextual prompt capabilities that enhanced rather than disrupted existing workflows.

My Role: UX/Product Designer working directly with engineering Timeline: 2 weeks Tools: Figma