Turning AI Deployment Friction Into Product Confidence (RHOAI)

Red Hat OpenShift AI is the platform enterprise teams use to build, test, and deploy AI models into real products. Red Hat wanted that deployment to feel like one click, so our five-person Human Factors capstone team partnered with them to find out why it didn't.

We interviewed 8 AI practitioners, 3 of them Red Hat engineers, and the same three problems kept surfacing. Failed deployments explained nothing, so engineers switched between consoles to piece together what went wrong. Technical terms went unexplained, pushing people out of the product to find answers. And hardware setup forced a choice between simple presets and the control experienced users actually needed.

We built a Figma prototype around three fixes: a troubleshooting sidebar that names the likely cause and the action to take, consolidated logs for every deployment state, and contextual tooltips that explain terms in place and link to documentation, plus hardware presets that can still be customized in full. We tested it across two rounds with 9 Red Hat engineers, designers, product managers, and RHOAI leadership, in person in Boston and then over Zoom.

The final version eliminated the console switching testers called their biggest problem. RHOAI leadership told us it supported the same things their own research had found over the previous two years. As one of two UX researchers on the team, I ran practitioner interviews, helped synthesize the findings, and ran the second round of user testing.