AI coding assistants excel at implementation, but they often inherit vague or incomplete requirements. This talk argues that the real bottleneck in AI-assisted software development is not writing code faster—it's making better product decisions before a single line of code is generated. Instead of asking agents to build immediately, we can first ask them to interview us, exposing assumptions, surfacing trade-offs, and turning implicit knowledge into explicit specifications.

Using a deliberately ambiguous product case study, we'll explore a progressive interviewing workflow where AI evolves from asking generic questions to acting as a product sparring partner that recommends options, explains trade-offs, and captures decisions as durable project memory. The result is a repeatable process that produces higher-quality specifications, shared team context, and a stronger foundation for AI-assisted implementation.

Attendees will leave with practical prompts and workflows they can immediately apply to transform coding agents from code generators into collaborative product interviewers—reducing ambiguity, improving specifications, and increasing the quality of every AI-assisted development project.

First delivered at Cursor Cafè, 2026, Milan, Italy