Luke Keller

Engineering + Compliance + Product + Infrastructure - Leading teams that don't report to you 🚀 Four people sitting in my office. All reporting to me. None agreeing with each other. We were six weeks from launching our fraud detection feature. The one that would make or break our quarter. Engineering: "It's ready. We can ship next week." Product: "Finally. Customers have been waiting. Let's go." Compliance: "Not so fast. This flags claims as fraudulent, but nobody knows why. Customers don't trust black box AI. We need explainability." Infrastructure: "And at 2,000 claims per hour, our architecture crashes. We need Lambda orchestration with Fargate containers, auto-scaling, proper state management. Right now we hit limits at 500 claims." Product: "So we miss our deadline for infrastructure upgrades?" Engineering: "Adding SHAP and LIME means rewriting the model. That's scope creep." Four smart people. Four valid concerns. All looking at me to pick a side. I didn't want anyone to lose. So I asked: "How do we make this work for everyone?" The energy shifted. Product: "Limited pilot. One customer. 50 claims per hour. See what breaks." Engineering: "At that volume, we can add explainability without worrying about performance yet." Infrastructure: "We learn the bottlenecks at low scale, then architect for 2,000 claims per hour based on real data." Compliance: "I can review every explanation the AI generates. Make sure customers actually understand them." Twenty minutes earlier, they were fighting. Now they were building something none could have designed alone. The pilot was messy. Infrastructure found Lambda cold starts killing response times. Engineering discovered SHAP calculations were more CPU-intensive than expected. Compliance caught explanations that were technically correct but impossible for customers to understand. But by the time we hit full scale, we had a Lambda + Fargate hybrid that Infrastructure built from real traffic patterns, explainable AI that Compliance refined with customer feedback, and performance optimizations Engineering discovered during the pilot. We signed both deals. And three more after that. The lesson: You don't lead by picking winners. You lead by asking "How do we ALL win?" Engineering fears technical debt. Product fears missed revenue. Compliance fears trust erosion. Infrastructure fears systems that collapse under load. Your job isn't deciding whose fear matters most. It's creating space where all fears get addressed and everyone contributes. The best products aren't built by one function winning. They're built when all four shape something nobody could have imagined alone. What's your framework for getting teams to solve problems together? #Leadership #ProductManagement #EngineeringLeadership #CrossFunctionalTeams #TeamAlignment #SHAP #LIME #ExplainableAI #LAMBDA

6