Guide

When to automate a workflow, and when not to

Not every process should be automated. The best candidates are frequent, rule based and painful. Here is the checklist we use in discovery calls.

Related service: workflow automation

By the UNIDEX team · Published 26 October 2026

Start with the work that repeats

Automation pays off through repetition. A task that takes ten minutes but happens forty times a week is a far better candidate than a complex task that happens twice a year. Before looking at tools, list the tasks your team does every day or every week and roughly how long each one takes.

The five questions we ask about every process

  1. How often does it happen? Daily and weekly work comes first.
  2. Are the rules clear? If two people would make the same decision with the same information, a system can usually make it too.
  3. Where does the data live? Processes that move data between tools you already use are the easiest to automate.
  4. What does a mistake cost? Low risk steps can run fully automatically. High risk steps get a human review step.
  5. Is the process stable? If it changes every month, document it and settle it first.

Good first candidates

  • Copying data between forms, email, spreadsheets and your CRM
  • Reminders, renewals and follow ups that depend on someone remembering
  • Weekly and monthly reports built from the same sources
  • Document generation from templates: proposals, contracts, invoices
  • Routing incoming requests to the right person

When not to automate (yet)

Do not automate a process nobody can describe. If every person on the team does it differently, automation will lock in the confusion. The same goes for work that depends on judgement, relationships or negotiation: there, automation should prepare and summarise, while the decision stays with a person.

Also be careful with very rare tasks. Building and maintaining an automation has a cost, and a task that happens a few times a year is often cheaper to keep manual with a good checklist.

Design for exceptions, not just the happy path

Most automation projects fail on the edge cases: the order with a missing address, the payment that does not match, the document in an unexpected format. A good workflow knows what it cannot handle and sends those cases to a person with all the context attached. That is the difference between automation your team trusts and automation they quietly work around.

A simple way to start

Pick one workflow, measure how long it takes today, automate it with a clear review step, and measure again after a month. One well chosen workflow builds the confidence and the data you need for the next one.

Questions

Good to know.

Which process should we automate first?

The one that is frequent, rule based and already moves data between tools you use. It gives a quick, measurable result.

Should every step be fully automatic?

No. Low risk steps can be automatic, while decisions with real consequences should include a human review step.

Let's build together

Have a workflow worth improving?

Whether you have a clear plan or just a starting point, we would love to hear what you are working on.