Refactoring Orgs

Two stories recur across engineering discussion threads. A request answerable with a few SQL queries becomes a large multi-month project because nobody asked the right question early enough. A team spends six figures in engineering time on something nobody needed, then leadership sells it as important rather than admit the waste.

Both stories have the same structure. Someone had the authority to build things. Nobody had the authority - or the incentive - to stop building the wrong things.

AI does not touch this problem. It accelerates the build cycle on both right and wrong things equally. An organization that systematically builds features nobody uses will do so faster with AI-assisted development. The output rate goes up. The waste rate goes up proportionally.

Feature factories don't fail because they build slowly. They fail because the stopping mechanisms - user research that actually blocks decisions, product authority that can kill work in progress, financial accountability for what gets shipped - are absent or deliberately weak. Velocity was never the constraint.

Speed makes this worse, not better. The waste compounds faster. The debt arrives sooner. The gap between what was built and what was needed becomes visible in months instead of years.

That's not nothing. But faster visibility into failure is a consolation prize, not a solution.