Use RPA to assist a business user when completing a BPM task

In the use cases so far, within the context of an orchestrated process, an activity is either completed entirely by a person or entirely automated by the system. BPM can also provide assistance to business users when they are part way through completing a task by integrating with some back-end system. For example: retrieving an address based on a specified Zip code, retrieving matching claims based on some search criteria, or using Watson Analytics to provide personality insights.

These integrations are typically implemented within a BPM human activity by creating an integration service and using synchronous system APIs. It is not appropriate to use RPA as an alternative technique to implement the integration service because RPA activities are asynchronous, so the time required to complete the request to the back-end system would probably be longer than the user would expect.

However, RPA can be used to assist a business user to complete their BPM task by integrating directly with the BPM task UI on the user’s desktop. This is a technique known as desktop process automation or attended robotic process automation. This is particularly useful for swivel chair activities, where the user has to log into multiple systems and transfer data between them and the BPM task UI. This pattern requires that the RPA robot resides on the user’s desktop, so that it can operate the user’s browser to interact with the BPM task UI. However, this may be impractical where you have many operators that each require their own dedicated RPA bot runner.

See sample: Use RPA to assist a business user when completing a BPM task

Note: Although this sample demonstrates how desktop process automation can be used with BPM, the best-practice approach for implementing swivel chair activities is to have the RPA bot runner deployed on a shared server, as described in: Use RPA to complete a BPM activity within an end-to-end business process.

Join The Discussion

Your email address will not be published.