Artificial Intelligence (AI) Meets Digital Duct Tape (RPA)

Insurers are beginning to explore AI as a potential enhancement to existing RPA tools or possibly as a complete replacement.

(Image credit: Lucas Dudek/Unsplash.)

Robotic process automation (RPA) has been used in the insurance industry since the 1990s and has incrementally improved over the years but still suffers from issues related to the underlying systems it must depend upon. Currently, the rigid and fragile nature of most RPA implementations leads to stalled processes and poor customer service. Moving from screen position field identification to tagging fields and incorporating API connections to RPA solutions where available has made the process better, but it still lacks the resilience and adaptability necessary to become the next digital duct tape.

Insurers have relied on RPA to bridge the gap between older insurance systems that contain a large amount of insurance capability but lack the process functionality to support a modern digital strategy with some success. Unfortunately, the requirements to meet the industry bar for good digital service strain the fragile RPA-to-older-system connection.

Insurers are beginning to explore AI as a potential enhancement to existing RPA tools or possibly as a complete replacement. There are some vendors working to use AI to deal with the need for greater resilience and adaptability in the RPA tooling but now, most insurance AI/RPA integrations are using AI capabilities as a step in a larger RPA managed process, generally around underwriting support and claims management.

Life, Accident & Health insurers need a reliable and adaptive RPA capability to maintain pace with accelerating consumer and employer digital service expectations. Most insurers ultimately must execute on core system replacement to suit new products and markets that cannot be serviced by incumbent systems, but the timelines for full system replacement require robust interim solutions now to remain competitive. The right combination of AI and RPA capabilities can be the basis for those interim solutions if insurers and vendors can take a more holistic approach to solving the problem.

Looking Deeper

RPA is not new to the insurance industry and has been used since the 1990s to do basic screen scraping of mainframe systems for user interface front-end modernization and basic scripting. These implementations were fragile and could be broken by a simple date format change. Over time, the RPA solutions have become more robust but still require significant upfront mapping work and lack process resilience.

Some ask why RPA is so important in a time when there is such focus on APIs and microservice architecture as the way forward. The issue is that much of the most robust and detailed insurance business logic an insurer needs is still held in incumbent systems that were not designed for granular access to APIs. It’s not a purely technical problem because one can build RESTful APIs to connect to Mainframe (zOS) COBOL and PL/1 applications, but there would be considerable work in the existing application to connect that API to internal business logic. In most cases, using RPA to manage the existing user interface to activate a process or extract data is a much quicker path to make that connection. It makes sense as long as the connection remains reliable and resilient to change.

There are a few companies that are working to use AI to make RPA tool more resilient and holistic like Deloitte’s work on Robotic and Cognitive Automation and SS&C Blue Prism. According to notes from a recent Datos CIO council meeting, there is a lot of interest in seeing if AI can replace RPA and manage the process itself as well as add another level of intelligence and decision making. Personally, it would be great to see these combined, but my concern would be whether it could scale for production.

AI already requires a large amount of processing power to build LLMs (Large Language Models) and then to query them. I worry about the ability to scale a generative AI driven daily workflow process for hundreds of simultaneous users and tens of thousands of transactions. The AI tech community is proposing the power requirements for generative AI and the need to be green require a resurgence in nuclear power generation. That’s not necessarily a bad thing but may generate additional AI controversy on top of the other concerns around information privacy and the risk of plausible but inaccurate results.

What’s Next

Over the next few years, generative AI will find its place as a powerful additional capability in the insurance technologists toolbox and will incorporate into the larger business/technology architecture supporting the evolving insurance market. It will become part of large core platform solutions, incredibly powerful analytical solutions, and a key component of the next generation of “digital duct tape.”  Any good engineer will tell you even the best designed, comprehensive platform of any kind will eventually need to connect to something that was never in scope of the design, or just spring a leak when it is stressed beyond original spec, and you will need duct tape.

RPA Can’t Think: Insurers Should Invest in Cognitive AI

Chuck Johnston // Chuck Johnston is an insurance business technology strategist and founder of Johnston Digital Ventures, where he provides advisory and consulting services for the Life, Accident and Employee Benefits space. He is a recognized expert in applying new business and technology concepts to the unique processes, architecture and technology of the insurance industry. With more than 30 years of experience, he is a frequent presenter at industry conferences, bringing a background in the insurance carrier, analyst, and software vendor communities. Chuck previously held leadership roles at Callidus Software (Callidus Cloud), Siebel, Oracle, Celent Research, EIS Group and FINEOS. LinkedIn:

Leave a Comment