Every operations team has the same conversation eventually: "we should automate this." The hard part isn’t deciding automation is a good idea — it’s figuring out which parts of a process are actually worth automating, and which ones will bite you later if you do.

Start with the process, not the tool

The most common mistake we see is picking a tool first — n8n, Zapier, Make, whatever’s trending — and then trying to force a process into it. Work backwards instead: map the actual steps a human takes today, including the judgment calls, the exceptions, and the "well, it depends" moments that never make it into the official process doc.

A workflow with a lot of judgment calls isn’t necessarily a bad candidate for automation — but it usually means you need AI in the loop, not just rule-based triggers, and that changes the tooling conversation entirely.

The three questions that actually matter

  • Is it repetitive? If a human does the exact same sequence of steps more than a few times a week, that’s automation territory.
  • Is the data structured, or can it be? Automating a process that runs on tribal knowledge and verbal handoffs means you’re really automating a data problem first.
  • What happens when it’s wrong? Every automation eventually hits an edge case. The question isn’t whether that happens — it’s whether your workflow fails loudly (a human gets alerted) or silently (nobody notices until a customer complains).
The goal isn’t to remove humans from the process. It’s to remove humans from the parts of the process that don’t need a human.

What we’d actually leave alone

Not everything should be automated, even when it technically could be. Anything involving a genuinely novel judgment call — a first-time vendor negotiation, an unusual customer escalation, a decision with legal exposure — is usually better served by a human with good information in front of them than a fully automated flow.

The workflows that age well are the ones where the automation handles the predictable 80% and hands off cleanly to a person for the 20% that needs real judgment — with full context attached, not a bare notification.

Where to start

Pick the process that’s costing the most hours right now, not the one that seems most "automatable" on paper. A messy process that eats 15 hours a week is a better first project than a clean one that only saves 30 minutes — even if the messy one is harder to build.