Automate the operations, not just the marketing

With the process that is most repetitive, most defined and most expensive in time. Repetition and clarity matter more than importance — an important process with fuzzy rules automates badly, while a boring well-defined one delivers immediately and builds internal confidence for the next.

Processes requiring genuine judgement, processes that change frequently enough that maintenance would exceed the saving, and processes that run rarely. Automating something that happens twice a year costs more than it returns and then quietly breaks between uses.

In our experience it changes what people spend time on more often than it reduces headcount. Coordination and data entry go away; exceptions, judgement and customer relationships remain. If the business case depends on removing roles, test that assumption carefully before committing.

Rarely the technology. It is that automating a process forces you to define it precisely, which surfaces disagreements about who owns what and when a step is complete. Those conversations are valuable and they are also why these projects take longer than the build estimate suggests.

By designing for them explicitly rather than pretending they will not occur. Every automated process needs a defined path for the case that does not fit — usually escalation to a person with context. A process that only handles the standard case fails visibly the first time reality intervenes.