I recently participated in a fascinating panel discussion, sponsored by Nuxeo, led by Insurance Post and entitled Insurance in the digital age – Transforming claims processing with AI-driven automation. Two senior Insurance veterans: Tina Winterburn, Claim Director at Gallagher, and Mike Cummings, Principal at Aon Inpoint Claims, joined me for this discussion. We covered a wide range of topics, from the technical challenges brought about over the past year by the global pandemic to the strategic investments that insurers are making to improve customer satisfaction, increase operational efficiencies, and protect their position against upstart InsurTech competitors.
This substantive discussion explored the many ways traditional insurers prioritize innovation to address recent challenges brought about by the global pandemic and new technologies already on their innovation agendas. While Artificial Intelligence (AI) was a significant theme for the discussion, the discussion centered on why these technologies were essential and how insurers were using them to improve their business’s critical aspects.
While I encourage you to watch the entire panel discussion, three takeaways are worth highlighting. Also, I’ll respond to a few questions from attendees that we could not address during the session.
Intelligent Automation, Fueled by AI, Drives Efficiencies
One objective that both Tina and Mike mentioned was turning over claims as rapidly as possible without increasing claims leakage. Mike Cummings highlighted that:
supervisors at Aon Inpoint Claims spend upwards of 35% of their day looking at files related to open claims to determine those processes’ status.
AI’s ability to infer essential information from these documents simplifies this process for supervisors and can streamline the claims processes. The system facilitates the supervisor’s tasks by improving the quality of information that is automatically gathered from these documents, improving supervisor efficiency. Instead of spending a third of their time analyzing these documents, supervisors can save considerable time by quickly identify the documents and claims processes on which to focus. Finally, team members can now identify these insights immediately, without relying on support from their supervisors.
Different Types of AI for Different Objectives
The discussion revealed several different objectives:
- identifying claims information
- inferring customer sentiment to improve responses
- flagging potential escalations that require dedicated attention
Understanding how different AI types can improve the claims processes is essential, as no single AI technology will address your needs. While the Nuxeo Content Services Platform includes sophisticated AI technologies, many of our customers layer their AI to address different business needs. For example, each of the most prominent Cloud infrastructure vendors provides AI services to perform specific tasks, including audio and video transcription, document translation, and sentiment analysis. These generic AI services handle everyday tasks that may apply to various use cases in businesses across different industries.
Learn more about generic vs business-specific artificial intelligence here.
But these technologies are difficult to tune to address particular requirements. Nuxeo Insight, our AI in content management service adds the ability to identify common elements across both text and multimedia formats. The business needs may be as simple as categorizing the incoming documents to identifying specific details in a photograph, such as the type of automobile or the extent of its damage. By automatically identifying critical aspects of these documents, AI can automate significant portions of the claims process while noting those documents or claims that demand closer inspection by the claim teams or supervisors.
Low Code Tools are the Key to Unlocking Complex Technologies
While the promise of AI is exciting, implementing AI has traditionally been a massive undertaking. A critical success factor is the talent and experience of the data scientists that built the AI models. Not surprisingly, large enterprises that invest heavily in AI achieve significant benefits. At the same time, organizations frequently complained that they did not get the expected results or that their implementations were much longer and more expensive.
But there is a better way. AI technologies that deliver low-code tools for building models lead to faster and more successful implementations. Nuxeo Insight, for example, automatically creates models based on sample documents. Based on the commonalities identified amongst a sample set of documents, Nuxeo Insight can deliver these insights for incoming documents automatically. By creating a low barrier to entry for AI, Nuxeo Insight unlocks what has often been complicated technology to get right.
While we answered several questions from attendees during the session, there were a few key questions to which I wanted to respond:
- What are the biggest barriers to implementing your ideal AI-powered solution? Is it the right technology, the right data, the right internal skill sets or a combination of all?
As I’ve already noted, the skill set required to implement AI is a significant barrier to successfully implementing AI solutions. The good news is that vendors are responding, and organizations are starting to see AI technologies, including Nuxeo Insight, that provide low-code options for creating AI models. By removing the need for expert skills, organizations can focus on the relevant AI technologies to deliver insights to drive intelligent automation.
- Mike suggests that we use the AI to support decision-making of experienced staff. Is there a risk that we go too far, and we rely on historical trend data which isn’t representative for changes in the claims world as cultures/technology evolves?
This question raises two essential facts regarding AI: to maintain accuracy, it is vital to regularly inspect AI models with the understanding that AI models are rarely “finished.” It’s common to get excellent results immediately after training an AI model. Over time, however, a more comprehensive range of documents can reduce accuracy levels. By regularly adding new document samples to the training set, organizations can maintain their models’ accuracy.
- Would there ethically be a way that as a claims handler I can understand how my customer is feeling without physically speaking with them?
We alluded to this earlier, but AI-driven sentiment analysis is a great way to gain insight into a customer’s feelings when communicating online or exclusively through a digital process. Organizations can use widely available AI technologies, including Amazon Comprehend, Google Cloud Natural Language, and Microsoft Azure Text Analytics, to react appropriately to their customer’s feelings.
While AI and automation offer significant value to insurers, implementing AI is not without challenges, driven mainly by scarce AI expertise and limited insights from generic AI services. Despite these obstacles, forward-looking insurers find that business-aware AI services, combined with low-code tooling, can enrich your content to support increased automation and faster processing with fewer manual tasks and increased accuracy.
Learn how to power your content with Artificial Intelligence in this whitepapper.