Automation is easy to buy and hard to regret slowly. A new tool looks efficient in a demo. Six months later the team is maintaining zaps, copying data between systems that still do not agree, and doing the original work whenever the automation fails.
The useful question is not ‘what can we automate?’ It is ‘where does work actually get stuck?’ Start with the work people repeat every day before choosing the tool.
Good candidates
Lead routing that today depends on someone noticing an email. Follow-up that dies if the salesperson is busy. Status updates that exist in three places and match in none of them. Reports that take a morning to assemble from exports.
These are repetitive, rules-based, and expensive when they slip. Connecting them usually saves hours and reduces the chance that a customer waits because a hand-off was missed.
Poor candidates
Judgement. Pricing exceptions. Sensitive customer conversations. Anything the team cannot yet describe in a sentence. If people disagree about what ‘done’ looks like, automation will scale the disagreement.
Tools also add complexity when they sit on top of a messy process. Automating a broken handover does not fix the handover. It hides it until a customer feels the delay.
A practical sequence
Map the work as it happens, not as it is supposed to happen. Count the hours and the drop-offs. Connect the systems that already hold the truth. Automate the steps that are clearly repetitive. Measure hours saved and whether customers move faster.
If those numbers do not move, the automation was decoration. Keep the parts that removed real friction. Remove the rest before it becomes another system the business has to live with.