The Challenge
New flow for a new experience
The large retail company’s onboarding flow asked a new client roughly seventy questions plus a fixed style quiz before it would say anything useful back, and a flow that treats a quick shopper and someone who genuinely wants to explore their style as the same person ends up serving neither one particularly well. What the company needed was a flow that adapted to the person answering it, rather than a form that assumed everyone wanted the same depth of interrogation before getting a recommendation.
The Approach
Personalization in real time
Solvd built a standalone web application around a chat-like interface with custom components, running on an agentic AI system that interpreted both the conversation and any reference images a client uploaded, extracted styling preferences from what was actually said rather than from a checkbox, and adapted the question flow in real time based on what it had already learned. From that same conversation, it generated personalized on-figure outfit visualizations grounded in the client’s actual available inventory, so what a client saw wasn’t a hypothetical outfit; it was something they could actually receive.
The Outcome
Human-centric improvements and business results you can’t ignore
The company piloted the flow with real customers before committing to a broader rollout. Optimization of the AI experience produced clear improvements: average session cost fell 25.5%, cost per conversational turn fell 27.2%, and session duration fell 15.7% versus an earlier version of the same solution. OpenAI models coordinated the underlying agents, interpreted both text and images, extracted the structured preferences that drove the flow, and adapted the onboarding journey in real time.
