01
AI-powered question suggestions and a built-in chat assistant
Research with our most active users showed they want AI to be a proactive partner when building, not just a tool that responds when asked. This project extends the existing chat assistant in the inspection builder with automated suggestions, and brings it together in a new, unified interface.
VIEW MORE
01
AI-powered question suggestions and a built-in chat assistant
THE PROBLEM
Users wanted a proactive building partner, not just a chatbot that responds when asked.
I led 8 research interviews, spanning over 10 hours, with our most active users of the inspection builder. The interviews were largely exploratory with one key goal: how could AI improve the way you build an inspection? Question suggestions quickly emerged as one of the top requests. Again and again, there was a clear sentiment from our users — they don’t want AI to build for them, they want AI to build with them.
I followed these interviews up with some data analysis on adding questions. Less than half of new sign-ups even create a question the first time they open the builder, and those who do take over 6 minutes on average. There was clearly work to be done here.
Opportunity and user journey map from customer calls
Data analysis of users adding questions
THE APPROACH
One panel, with AI chat and suggestions together in one place.
We chose to zero in on AI question suggestions. It’s a relatively simple feature with enormous potential benefit, giving theoretically the greatest return on investment.
The inspection builder already had an existing AI chat function, but in a floating window that didn’t feel embedded in the interface. To add question suggestions, I knew the two had to be combined into one homogeneous experience.
Beyond the UI, I also had to sort out where the suggestions came from. With several ways into the builder, I mapped how data flows from input, to the AI engine, to the suggestion output.
UI refinement of trigger points
Map of how AI pulls relevant context based on entry flow
THE OUTCOME
~5k suggested questions were added to inspections in the first 30 days.
In the first 30 days, we saw 8.5k clicks on the question suggestions tab with over 5k questions added, equating to roughly one question for every two visits. Paid users were the heaviest adopters, demonstrating value for our paid customer base.
