"Automate everything" is one of the most repeated, and most confidently stated, pieces of business advice in the automation and productivity space, and it's wrong often enough in practice that it deserves genuine pushback rather than repetition. The advice sounds efficient and forward-thinking in the abstract, but applied literally, it produces predictable, avoidable problems that a more selective approach avoids entirely.
It ignores that build and maintenance cost real time and money
Every automation, however small, costs time to build and ongoing time to maintain, monitor and eventually fix when something inevitably changes upstream. Applied to a low-frequency, low-value process, that cost frequently exceeds whatever time the automation ever saves in practice, making it a net loss quietly dressed up as an efficiency win.
It treats all processes as equally suited to automation, when they aren't
Processes requiring genuine, case-by-case judgement don't automate well regardless of effort invested, and forcing automation onto them tends to produce either an oversimplified rule that gets edge cases wrong regularly, or an increasingly complex system trying and failing to capture every possible nuance a human would have handled naturally and without incident.
It underweights the maintenance burden that compounds over time
Ten small automations, each individually justified in complete isolation, become a genuinely significant maintenance burden once viewed collectively, especially once the person who built them has moved on to other things and nobody has full visibility into how they all actually work together as a system. Blanket automation advice rarely accounts for this compounding cost across an entire portfolio of automations.
It can remove valuable human awareness embedded in manual work
A person doing a task manually often notices related problems, an unusual pattern, a customer who seems off, context that matters, simply by being present and paying attention in a way a narrowly scoped automation never will. Automating away all human involvement in a process can quietly remove this kind of ambient awareness along with the intended, targeted efficiency gain.
A better mental model: automate what clears the bar, not everything
The useful question was never whether to automate, it's which specific processes genuinely clear a reasonable bar on time saved, complexity, and risk, discussed at length elsewhere in this series, and in what order to tackle the ones that do. Some processes should stay manual indefinitely, and that's a legitimate, considered outcome of a proper evaluation, not a failure to automate aggressively enough.
What to actually do instead of chasing full automation
Score every candidate process honestly against time saved, complexity and risk. Automate the clear winners first, without exception. Revisit the ones that didn't clear the bar periodically as circumstances genuinely change, rather than either forcing them through prematurely or dismissing them permanently. This selective, evaluated approach is what separates automation that genuinely serves a business from automation pursued as an end in itself, applied everywhere simply because it's technically possible to do so.
The discipline this requires from a team
Saying no to automating a process that doesn't clear the bar takes more discipline than saying yes to automating everything in sight, since the latter feels productive in the moment even when it isn't. Building the habit of running the actual evaluation, rather than defaulting to automation because the tools make it easy, is what separates automation that genuinely serves a business from automation pursued for its own sake.