AI won’t save a product that was never composable

AI won't save a product that was never composable Most product teams shipping AI right now are following roughly the same playbook. A chat sidebar. A summariser. A copilot panel that can answer questions about whatever's on screen. Each release gets a launch post. Each one tests reasonably well in isolation. And yet, a year in, something is off. The product feels louder, not better. Customers use the AI features once or twice and go back to doing things the old way. The team keeps shipping. The needle doesn't move. This is not an unusual story. It's close to the median story.