2-day virtual workshop
Offered three times in 2026; select your dates when registering:
Option 1: May 4 and 6, 2026, 8:00am-12:00pm PT
Option 2: June 30-July 1, 2026, 8:00am-12:00pm PT
Option 3: October 30 and Nov 6, 8:00am-12:00pm PT
Somewhere between 50 and 80% of AI projects fail. That’s at least double the failure rate of traditional IT projects. Companies have poured tens of billions into AI, and the returns so far are dismal: one recent MIT report found 95% of organizations getting zero return on their investment. The problem usually isn’t the technology. It’s that teams keep picking the wrong projects: technically ambitious, high-risk endeavors that demand near-perfect accuracy to be useful, when what most organizations actually need are the many modest, valuable, low-risk AI projects hiding in plain sight. The traditional user-centered design process alone won’t find them.
This workshop teaches a set of techniques that designers, product managers, and researchers can start using immediately. It adapts what we’ve been teaching our graduate students at Carnegie Mellon’s Human-Computer Interaction Institute for working practitioners: short lectures to introduce each method, then hands-on exercises to practice it. You’ll work in teams on real concepts, not hypotheticals.
Part One is about building the right thing. When you can build anything, how do you know you’re building something valuable? We’ll start with AI Capabilities: learning to see AI not as a monolith but as a wide range of dead-simple abilities (sort items into categories, detect objects in images, predict an event) that can be matched to real human contexts. Then Matchmaking, a brainstorming technique for systematically pairing those capabilities with organizational needs to generate concepts. You’ll rapidly evaluate those concepts for technical feasibility and value, run them through Consequence Scanning to surface unintended harms before you build them into the world, and finish by ranking your concepts to find the ones actually worth pursuing.
Part Two is about building the thing right. Designing for AI means designing for ambiguity, in both what users ask for and what the system does in response. We’ll practice Intent Mapping: connecting what users actually want to what the AI does and what users do next, and deciding when a concept needs a conversational UI, a traditional GUI, or no UI at all. We’ll explore Adaptive UIs, which find the frequent, repetitive moments in existing products where AI can quietly save users time or personalize the product. Then Human-AI Collaboration: designing systems where people direct, monitor, intervene, and approve rather than blindly accept whatever the machine produces.
We’ll end with Explainability: how much users need to know about how the AI works to trust it appropriately, and when an explanation does more harm than good.
You’ll leave with methods you can run with your own team the next day and a sharper sense of which AI projects deserve your time and tokens. Most don’t. The good ones are worth finding.
