I’d categorize RPA workloads into two broad categories:
- Scheduled or pre-planned workloads. This would be easiest for an AI to target.
- Ad-hoc workloads
Improvements could happen at:
- The process execution (i.e. improvements to the process to make it use less CPU, RAM, improve the execution speed)
- The process outcome (i.e. changing the process itself, reordering process steps, improving or retraining any AI systems inside the process, designing new processes)
- The scheduling level (i.e. changing the schedule to smooth out workloads and improve SLAs. Prevent work items from waiting in their queues)
- Process to Computer assignment level (i.e. improve utilization of computers)
- The number of robot workers (i.e. scaling up and down the number of computers)
I’ll try to look at how AI can affect each one of those different parts in further posts. I haven’t given them enough thought yet.