The automation maintenance tax nobody budgets for

The pitch for automation is almost always framed as a one-time cost: build it once, save the labor forever. In practice, every workflow shipped is a small piece of infrastructure that needs upkeep, and the businesses that skip budgeting for that upkeep are the ones who, eighteen months later, have a graveyard of "automations" nobody trusts anymore.

The maintenance tax shows up in predictable ways: an upstream API changes its response format without warning, a spreadsheet column gets renamed by someone who didn't know an automation depended on it, a service the workflow calls gets deprecated. None of these are edge cases, they're the normal lifecycle of any system with external dependencies.

The businesses that avoid the graveyard treat automations like software, not like a one-off project: someone owns each workflow, a lightweight alert fires when a run fails silently, and there's a quarterly pass to check that nothing upstream has quietly drifted out from under it.

Budgeting five to ten percent of the original build cost annually for this upkeep isn't pessimism. It's the difference between an automation that's still running correctly in year three and one that's been quietly broken for the last four months, with nobody noticing until a customer complains.