Notes on work · July 20, 2026

Why do most automation projects fail?

Nobody tells you about the abandoned ones.

Every business I talk to has one: an automation or an AI tool somebody bought or built, used for three weeks, and quietly stopped using. The subscription is still active. The spreadsheet it was supposed to replace is still open in a tab.

The failure almost never happens in the software. It happens earlier and later.

What goes wrong before the build?

Most automation fails because someone skipped the understanding. The project started from a demo, not from the work. Nobody mapped the workflow as it actually runs, so the system was built for the org chart's version of the process, and the org chart's version doesn't exist on a Tuesday afternoon. The person who really operates the workflow was never interviewed. The judgment calls hiding inside the process were never separated from the repetition, so the tool was asked to make decisions it had no business making.

You can't automate work you don't understand. You can only complicate it.

What goes wrong after the build?

The first version is never quite right, and pretending otherwise is how automation ends up abandoned. Real work reveals what the plan missed within weeks: the edge case nobody mentioned, the step everyone forgot because it lives in one person's head. If nobody is watching the system run and adjusting it with the people who use it, every one of those small misses becomes a reason to go back to the old way. Trust in a new system is spent quickly and rebuilt slowly.

That's why we stay through the messy first weeks. Not because the build was wrong, but because the first contact with reality is part of the build.

How do you keep an automation project from failing?

Understand first: map the workflow with the people who run it before choosing any tool. Build small: automate the repetition, keep the judgment human, and work inside the tools you already run. Then refine on purpose: schedule the adjustment period instead of hoping you won't need one. When we leave, you own the system, understand it, and can change it without us in the room.

If you've got one of those quietly abandoned tools in your business, you're in good company, and it's fixable. Tell me where the work got stuck. That's what a workflow conversation is for.

Book a workflow conversation

Start with the work, not the software.

Book a workflow conversation