Large language models and generative AI are reshaping how we interact with software and how we work, yet integrating these technologies into real products remains unexpectedly difficult. Despite the proliferation of “AI features” in modern applications, many feel unreliable, unintuitive, or disconnected from real userneeds. Why is it so hard to make AI genuinely useful?

This talk examines practical lessons learned at Dropbox from two perspectives: how AI has reshaped our day-to-day engineering practices, and what we encountered while building Dropbox Dash—a system designed to help people find, organize, and connect their content using fast-evolving AI capabilities. We will discuss the technical challenges, shifting constraints, and emergent patterns that arise when developing products on top of rapidly moving AI infrastructure, and explore what it really takes to build AI that works.

Rene W. Schmidt is a Senior Principal Engineer at Dropbox, Inc. He works on Dropbox Dash, an AI teammate that helps people find, organize, and connect their content effortlessly, supporting Dropbox’s mission to simplify how teams work together. Previously, Rene worked at Uber, VMware, and Sun Microsystems. He spent 10 years in Silicon Valley, founded the VMware, Uber, and Dropbox engineering offices in Aarhus, and holds computer science degrees from both Aarhus University and the University of Washington.