Instead, I trained myself to start with questions. Just four this time, but they reshaped everything:
Why are users unable to quickly find the right support content on their first search attempt?
We found that many queries didn’t match our internal terminology, and our search logic wasn’t forgiving to vague or conversational phrases.At what point do users abandon the search?
Session replays showed users often refined their query once or twice, then dropped off — usually within 10–15 seconds of scanning results.What makes a search result feel trustworthy or helpful?
Users gravitated toward results with clear headlines, familiar product terms, or content with ratings and timestamps — things that signal “fresh and relevant.”Would AI-generated suggestions improve their experience or create new friction?
We tested other websites that have AI functionality in their search. The AI suggestions showed potential, but when not explained clearly, they created confusion — especially for users who were looking for specific, human-written help articles.
We’re still deep in the ideation phase, but asking these critical questions early on has already brought clarity to a once-vague problem. It’s helping our team move beyond surface-level fixes and dig into the deeper reasons behind user behavior, why people abandon search, what makes them trust a result, and where AI might support or complicate their journey.
This approach is not about rushing to implement flashy features. It’s about designing with purpose, grounded in evidence and empathy. We’re now exploring ideas that align more closely with real user needs, and this shift has created better conversations within the team, between designers, product managers, and even developers.
We believe this mindset is leading us in the right direction. While we’re still shaping our solutions, the patterns emerging from these early questions are promising. They’re not just guiding our design, they’re building confidence that whatever we create next will be more relevant, usable, and impactful.