The relocation trap
Many automation projects do not eliminate work — they move it. The team stops copying data between systems and starts maintaining the automation that copies it: fixing broken runs, checking edge cases, cleaning up after silent failures. The manual work changed shape, but the hours remained.
Automate the process, not the mess
The trap almost always comes from automating a process as it currently exists, mess included. Before any workflow is automated, it deserves the same question an engineer asks of any system: should this step exist at all? Simplify first, then automate what remains. Automated chaos is still chaos — only faster.
Design for failure from day one
Real-world automation runs in a world of expired API tokens, changed formats and unavailable services. Robust automation is designed for that world: it retries intelligently, alerts when human judgement is needed, and fails loudly instead of silently. The measure of quality is not what happens on a good day, but what happens on a bad one.
The result worth aiming for
Good automation is quiet. It removes categories of work so completely that after a few months nobody remembers the work existed. That is the standard worth engineering toward — not a dashboard full of running workflows, but a team whose time returned to the decisions that actually need people.