AI Makes Building the Wrong Thing Easier

AI coding tools have compressed software development timelines from months to days. A department team can build and deploy an application in a fraction of the time it once took. While development is fast, shallow discovery becomes an expensive mistake. Teams that skip up-front research such as stakeholder interviews, workflow observation, definition alignment, and structured requirements documentation wind up building the wrong thing at unprecedented speed. The results are tools that demo well, but fail in real institutional processes because they function without tribal knowledge and mishandle undocumented edge cases. This session draws on direct experience with building AI-accelerated applications for education institutions, including student placement optimization tools, data intelligence platforms, and IT support systems. In every engagement, the discovery phase surfaced critical gaps: processes that existed only in people's heads, terms that meant different things to different departments, and workarounds that had quietly become the real system of record. Proper methods of discovery make the invisible visible and enable the technology to optimize functioning systems. Attendees will walk away with a practical framework for a discovery scope before any AI-accelerated build, how to spot the gaps between described and actual workflows, and why cross-domain understanding on the building team is the single highest-leverage investment in any technology project.