Is this process ready to automate? A checklist

Some processes automate cleanly in weeks. Others turn into months of exceptions because the process itself was never settled. Eight questions to ask before you spend anything, and what to do when the answer is no.

Automation projects rarely fail because the technology does not work. They fail because the process being automated was never quite agreed: three people do it three ways, the exceptions live in someone's head, or nobody can say what a correct result looks like.

Before we build anything, we ask the same eight questions. You can ask them yourself.

The eight questions

1. Can you describe it in ten steps? If the process cannot be written as a short list of steps that the team agrees with, it is not settled enough to automate yet.
2. Does everyone do it the same way? Ask two people who do the work to describe it separately. If the descriptions differ, decide which is right first.
3. Do you know the exceptions? List the cases that do not follow the normal path. You will not find them all, but you should find the common ones.
4. Can you tell a right result from a wrong one? For an order, an invoice or a report, what does correct mean? If nobody can say, nobody can test the automation.
5. Is the data clean enough? Customer lists, part numbers, prices. Automation checks against these, and duplicates or gaps become errors.
6. Does it happen often? Daily or weekly processes pay back quickly. Quarterly ones rarely justify the effort.
7. Is there an owner? Someone who will notice when the automated version is wrong, and has the authority to decide how it should work.
8. Can the systems be reached? The systems on either side need a way to get data in and out: an API, an import, or at worst a reliable export.

If nobody can say what correct means, nobody can test the automation.

What to do when the answer is no

A no is not a reason to give up. It is the first piece of work.

If the process is not agreed, spend an hour with the people who do it and write down one version. That conversation alone often removes a surprising amount of waste.

If the exceptions are unknown, keep a simple log for two weeks: every time something does not follow the normal path, one line saying what and why.

If the data is messy, clean the part the automation will touch, not all of it. A customer list with the top two hundred customers correct is enough to start.

If there is no owner, find one before spending anything. Automation without an owner drifts out of step with the business within months.

A rough score

Seven or eight yeses: ready. A first phase can usually start straight away.

Five or six: nearly ready. The gaps are normally closed in the first week or two of a project, as part of mapping the process.

Four or fewer: fix the process first. Automating it now would mean automating the confusion, and you would pay for it twice.

• It can be described in about ten steps the team agrees with
• Everyone does it the same way
• The common exceptions are known
• A right result can be told from a wrong one
• The data it checks against is clean enough
• It happens daily or weekly
• It has an owner who will notice when it is wrong
• The systems on either side can be reached

Questions people ask

What if we score four or fewer?

Fix the process first: agree the steps, name an owner, clean the data. Automating it now would automate the confusion.

Which process should we score first?

The one your team complains about most, that happens every day.

Score your most painful process. If it comes out at five or more, tell us about it. If it comes out lower, we are still happy to help you settle the process first; sometimes that is all it needs.

Canonical URL: https://ugibo.com/blog/is-this-process-ready-to-automate