The focus for any RPA project should can this be automated? Is the task at hand repeatable and rules-based enough to function with little to no user intervention? Without fully understanding the task at hand and it’s caveats, etc. the project will surely fail.

By Marisa Brown – RPA is not a plug-and-play tool and must be set up with care. With a thorough understanding of how end-to-end processes work, organizations will need to select appropriate processes or activities for automation using standardized criteria. Evolving, emerging or immature processes—and those with large ratios of exception management—will not be good candidates for automation. More broadly, sophisticated RPA applications will require ongoing adjustment, and procurement roles may need to be restructured in order to fully realize productivity gains from RPA.

While it completes routine work faster and at a higher quality, RPA can also make mistakes faster if given bad data. Poor data quality or the insufficient definition of business rules can lead, for example, to RPA ordering the wrong parts fast and in large quantities. The quality of RPA, especially when incorporating artificial intelligence (AI) capabilities, will largely be a product of the maturity of the organization’s data management efforts.

Another word of caution is in order — there is a point of diminishing gains and even productivity losses depending on the number of bots that procurement is using. It makes sense that a procurement team would add additional bots as it grows its RPA capabilities. For instance, 35% of organizations that are optimizing the use of RPA have more than 20 bots. Yet APQC finds that gains from RPA begin to reverse at this point. For example, organizations employing more than 20 bots have more FTEs ordering materials and services and process fewer purchase orders per FTE than those that use 6-20 bots. Similar productivity losses occur for those organizations that have bots operating but not necessarily optimized. Read On: