I’ve noticed something when using AI for ideation: You can get 10 polished ideas in seconds… and sometimes they’re all the same idea wearing different hats. The same problem happens with research. Ask AI: “What are our users’ biggest pain points?” You’ll get 20 answers in seconds. But ask yourself: Where did those answers come from? Now give AI 400 real support tickets and ask the same question. Suddenly, you have something much more valuable: Same AI. Same question. A completely different level of confidence. That’s why I follow one rule when working with AI: AI processes evidence. It doesn’t create it. Here’s how I’d run a discovery sprint with AI in 5 days: 1️⃣ Build the evidence base Bring together reviews, support tickets, analytics, session recordings, previous research, and old decks. Don’t start with AI. Start with reality. 2️⃣ Find the patterns Give AI your actual research and ask it to identify themes. But make it trace every finding back to a source. No evidence. No insight. 3️⃣ Talk to real people Talk to 5 users. Let AI help create the interview guide, organize transcripts, and surface potential patterns. But you still have the conversation. AI can process the interview. It can’t build the human connection for you. 4️⃣ Synthesize Analyze a couple of interviews yourself first. Then use AI to help scale the synthesis. And pay attention when you disagree with it. Those moments can reveal assumptions, missing context, or insights worth exploring. 5️⃣ Ideate Now bring AI into the creative part. Don’t ask: “What’s the best solution?” Ask it to generate multiple directions. Challenge assumptions. Explore alternatives. Look at the problem from different angles. Then bring those ideas back into your design process. By Friday, you could have 3 evidence-backed concepts instead of 30 AI-generated guesses. The point isn’t to use less AI. It’s to give AI better inputs. AI didn’t replace the research. It removed the excuse for not doing it. Wish you all the best