Most companies play AI on hard mode. Execs try to be heroes and build out flashy new genAI products to get press and look sick in front of peers. But rarely is the splashy AI product the best way to nudge your company in the direction of AI-first. Here’s the progression of AI adoption I’ve seen work: Phase 1: Single player tools - Train your people on ChatGPT for use cases specific to their work - Train your engineers on Claude/Codex to increase throughput Phase 2: Single player processes - Rebuild a process dedicated to one task and one individual (i.e responding to support tickets) Phase 3: AI product features - Enhance a non-AI-native product with a genAI use case (I.e Notion AI chat feature) Phase 4: Multiplayer processes - Rebuild a complex process that requires adherence across many people and many steps (end-to-end sales process) Phase 5: AI-native product - End-to-end AI build that couldn’t have existed before the genAI revolution (I.e Gamma for slides)