What do robotic process automation (RPA) and artificial intelligence (AI) have in common? RPA vendors are getting billion-dollar valuations and talking more about AI, machine learning (ML), and cognitive computing.
RPA is not an emerging technology per se. It’s a modern update to an existing technology approach that has become a tool for insurers to bypass manual processes with a layer of automation and lightweight integration. AI, ML, and cognitive computing, on the other hand, are components of a true emergent technology category—one that is distinct from RPA, but which vendors can use to make RPA more intelligent and valuable.
At the most basic level, RPA is the automation of manual tasks or manual integration points with algorithms that use existing user interfaces rather than APIs or Web services. Modern RPA is a straightforward “screen scraping” approach that is easy to create through visual tools and has strong orchestration and scalability surrounding it.
So where do AI and cognitive computing come in?
An RPA algorithm can be pretty “dumb” (a souped-up macro, in essence). For example, an insurer could code an RPA tool to read data from a submission portal and then to enter or rekey that data into another an UW system. If that’s all it does, it still can be a valuable way to replace manual efforts and allow underwriters to focus on underwriting rather than excess typing.
An RPA algorithm can evolve from there to have a level of intelligent fault tolerance that makes it more than just a macro. If the RPA has been coded to read a submission screen and then the submission screen is redesigned, this more fault-tolerant version can figure out where a critical text box moved without requiring recoding. It’s possible the vendor uses some AI technology on top of the RPA system to help with that fault tolerance.
An RPA algorithm can begin to utilize AI to augment learning and to slide more seamlessly into an insurer’s existing processes. An insurer might install the RPA/AI system onto the servers where humans perform tasks instead of coding the RPA integration points; the AI technology can watch existing interactions and determine how to replicate that behavior. The RPA technology is still the piece that recreates the manual processes against existing user interfaces; it’s the AI technology that watches and learns the process itself.
Finally, an RPA algorithm can use AI and cognitive computing to make decisions. Pairing the two can enable RPA to decide how to handle errors or to make risk decisions based on an incoming submission. This final use case seems to blur the lines between RPA and AI, but there’s still an important distinction between the functionality of each. AI makes intelligent decisions, not RPA. RPA is what executes any AI decisions into the existing user interface.
Such AI-driven intelligent decision-making would still be valuable to an insurer even without RPA. Without RPA, an insurer might have cognitive computing/AI algorithms make risk decisions and then leverage those decisions to influence underwriting workflow or automation by providing suggestions in the UW workbench or via direct workflow integration. RPA comes into play when interactions with insurer systems sit on top of those systems, using the same user interfaces a human would use.
Cognitive computing can make RPA behavior smarter in the same way it can make built-in UW workbench automation smarter.
The most viable long-term solution for insurers is to integrate cognitive computing decision-making via APIs and Web services rather than via RPA. (This is true of most RPA use cases.) However, it’s better to utilize analytics and intelligent decision-making with RPA than to not use it at all.
For more on current adoption and deployment rates, use cases, and carrier interest in AI, RPA, and other technologies, see Novarica’s Research Council Study, Emerging Technology in Insurance: AI, Big Data, Chatbots, IoT, RPA, and More.
Add new comment