Process automation: where to start in a business

Key points

• Do not automate an unclear process.
• Choose high impact and controllable risk.
• Keep human review for exceptions.

Identify the right work

Inventory the repetitive tasks: copying data, preparing quotes, notifications, stock checks and reporting. Note the frequency, the duration, the error rate and the dependencies.

Calculate the priority

A good starting process occurs often, follows clear rules and uses accessible data. Estimate the monthly hours saved and the cost of the errors avoided.

• Volume and frequency
• Time per execution
• Cost of a mistake
• Number of exceptions
• Data quality

Design the controls

Define what happens when data is missing, when a system does not respond and when approval is required. Logs and alerts are part of the automation, not optional features.

Measure after launch

Compare the time, the errors and the response time before and after. Expand only after the limited process is stable and people understand it.

Example: from request to quote

A request received through a form can be validated, classified and sent automatically to the right person. The data can enter the CRM, and the client can receive an immediate confirmation. The team steps in for verification and preparing the quote.

Automation reduces copying and response time, but it should not send a complex quote without review if the information is incomplete.

The simple formula for priority

Calculate the monthly time consumed, the hourly cost and the average cost of errors. Compare the result with the implementation and maintenance effort. Frequent, stable processes usually end up first on the list.

Add a risk assessment. A large saving does not justify full automation if an error can produce losses or legal problems.

• Hours saved
• Errors avoided
• Reduced response time
• Operational risk
• Maintenance effort

Team adoption

People must understand what the automation does, how to verify the result and how to report an exception. Involve the users from the design stage and test with real situations.

A simple, visible and documented workflow is more valuable than a sophisticated automation that nobody can explain.

How to document the process before automating

Follow one real case from start to finish. Note the trigger, the data used, the people involved, the decisions and the result. Repeat the exercise for a typical case and for several relevant exceptions. The resulting diagram must be understandable to the people who do the work.

Separate the real rules from the habits created by the limitations of the current tools. Do not automate a useless step just because the team has been doing it for a long time.

Control after launch

Build a dashboard with the number of executions, the errors, the cases sent to a person and the time saved. Alerts must indicate what happened and who can intervene. A generic error message just moves the work into another form.

Review the rules when the process or the connected applications change. Automation is not a project you can forget after launch.

• Execution log
• Explicit alerts
• Owner
• Manual fallback procedure
• Periodic review

Frequently asked questions