Challenge
When introducing conversational LLM capabilities for the MVP, the team diverged on how AI should enter the product: through more flexible, distributed interactions, or a more centralized and stable chat structure. The disagreement came from high-impact stakeholders with conflicting expectations, and remained at the level of preference and expression, making it difficult to converge.
Decision
Reframed the disagreement from preference into comparable trade-offs, making each option’s position and applicable scenarios explicit. With no usage data available, prioritized a stable and observable entry point to learn how users actually engage with AI, including what they ask and how they act, so real needs could be learned from use rather than over-designed based on assumptions.
