Northwell Health

New technologies and solutions generally identified as Artificial Intelligence (AI) are appearing nearly every day it seems. The topic elicits a wide range of responses ranging from fear and skepticism to hopeful enthusiasm and certainly many others in between. As with any new technology there is good reason for all of these initial reactions and perspectives which I believe are very healthy when they exist together in leaders and teams that are charged with guiding strategies to develop and implement solutions. While the possibilities are tremendous, it can be challenging to identify an appropriate entry point that satisfies the desire to move forward and provides an acceptable risk level at the same time.

ROBOTIC PROCESS AUTOMATION

Robotic Process Automation (RPA) is an option that many industries are choosing as that entry point. This targets high volume repetitive processes and tasks that are largely transactional and rule based. Identifying processes or tasks that have very low exception rates and infrequent process changes are generally the best candidates and have a lower risk of error. These types of tasks or processes can have a high potential to reduce the time to complete tasks, lower labor costs and improve customer experience, especially where there is digital data capture. Automation in these instances can enable deployment of human capital to higher level tasks.

 While RPA and AI hold a lot of promise, there is much to consider when selecting use cases. The data that is required to fuel these solutions may have utility far beyond the identified use cases which in turn may be more valuable than the initial use

EXAMPLE USE CASES

In retail there are solutions being deployed to make product recommendations based on data captured during online interaction, browser history or online spending. This is frequently seen in online retail but is also supporting human workflow both telephonically and in stores. Banks are using online transaction monitoring and triggering alerts to customers based on preferences for communication, level and type of transaction and other rules that can vary by institution. Several service industries are automating booking of services and appointments as increasingly customers want to be able to manage these types of activities on smart devices including scheduling appointments with their doctor. There are other more back office like functions that are possibilities as well. In healthcare, denial and referral management workflows are good candidates as they can assist with routing of transactions or claims adjudication. Help desk functions are being automated in order to aid in routing tickets to appropriate queues for resolution or escalation by applying rules based on status or incident type. Customer support and feedback is a functional area where RPA is being used across industries as well. The more widely used solutions that I have seen to date are chat bots. These are primarily used to elicit feedback on customer experience following the receipt of services. They involve a sequence of simple questions sent to a device and require yes or no or numeric ratings that either result in an additional question or provide acknowledgement that feedback was received and appreciated. I have also seen this being used to manage prescriptions by local pharmacies. While these are measured and carefully thought out solutions and good entry points to AI, these types of solutions can also fall under what I consider risks of automation.

RISKS TO CONSIDER

In general, the direct interaction with a customer is one to pay careful attention to as there is a desire for ease of access but also a desire to interact with and be treated as a human being. RPA solutions are very literal as they are rules based. If you have received a call from an automated call center you may not have noticed that it wasn’t an actual person but asking an unrelated question like “what is the weather like where you are?” will expose that you are not speaking to an actual person. The rules that drive these bots can certainly be refined and trained over time but they cannot account for all scenarios of a human encounter. Perhaps more important is that when one picks up the phone to make an inquiry, there is still an expectation of human interaction.

While RPA and AI hold a lot of promise, there is much to consider when selecting use cases. The data that is required to fuel these solutions may have utility far beyond the identified use cases which in turn may be more valuable than the initial use. This is something to be very mindful of as you make decisions on whether to develop capability internally or choose a development partner. The intellectual property contained in your data is likely more valuable than the early solutions that you may be considering.