Which process should you automate first?
Frequency, exceptions and the cost of a wrong choice: three questions for choosing a first bounded intervention.
Redazione SYSTEM ·
Choose an activity you can describe
“Let’s automate the sales side” is too wide a scope. “Every new complete enquiry creates a contact in the CRM and assigns an owner” defines an entry, an action and an observable result.
Write one normal example before choosing the tools. Which data arrives? Who is supposed to receive it? Which state says the work has been completed?
Count the repetition, not only the minutes
A short but frequent activity can interrupt work dozens of times. For one week, collect the number of cases, the time taken and the corrections required. Separate execution time from waiting time: they are two different problems.
Put the exceptions on the table
What happens when a field is missing, the contact already exists or the destination service does not answer? A first useful workflow has a way to recognise these cases, keep state and involve a person.
If every enquiry requires a new decision and you cannot explain what makes the result correct, start from defining the process, or from a tool that supports the person.
Weigh the cost of a wrong choice
Preparing a summary to be reviewed and sending an offer to a customer have different consequences. Define which actions can be automatic, which require review and which stay with the team.
AI can help interpret variable text. A deadline, an assignment rule or a transfer between fields do not necessarily require a model.
A one-page brief
Before starting, collect:
- The event that starts the process and the source of the data.
- The action to complete and who is responsible for the result.
- Frequency and time observed today.
- A normal example, an incomplete one and a duplicate.
- The expected behaviour if a service stops.
- The actions that require human review.
- The indicator to compare after the intervention.
Test the first workflow on a small, observable scope. The next step depends on the exceptions that emerge and on the work the team still has to do.