Automation
Automation uses machines or software to carry out work with less direct human intervention. It can handle part of a process or the whole process, using fixed rules or learned behavior.
An invoice workflow can match an invoice to a purchase order, check the amount and route exceptions to a person. A fixed script can do those steps without artificial intelligence. A model might help extract fields from differently formatted invoices, but it is only one part of the automation.
Automation has degrees. Software may collect information, suggest a decision, wait for approval or execute the decision itself. The National Aeronautics and Space Administration (NASA) separates these functions in its discussion of human–automation interaction because automating information gathering does not mean a system should also control the final action.
That matters when measuring value. Count the human effort remaining, including review, corrections and exception handling. A workflow that processes invoices faster but creates costly payment errors has improved activity without necessarily improving the outcome. Give failures a clear destination and preserve a record of what was done.
Sources
- NASA: Research on Hazardous States of Awareness and Physiological Factors in Aerospace Operations — Pages 1–2 of the report discuss automation, levels of human involvement and four information-processing functions.
Go deeper
- How do you measure whether AI is actually paying off? AI, decoded · How to Measure Whether AI Is Actually Paying Off
- What's the difference between human in the loop and human on the loop? AI, decoded · Human in the Loop vs. Human on the Loop
- Microsoft Learn: Power Automate workflows docs
Explore practical app-to-app workflows and the environment they run in.