Speaking the Language: Conversational Platforms in Healthcare
Significant advancements related to conversation platforms and related technologies have occurred over the last several years. New avenues have provided opportunities for businesses to serve customers more effectively and also reduce their operating costs. Conversational technology has seen increasing adoption across a wide range of industries including banking, telecommunications, healthcare, and finance.
Healthcare Business Landscape
Health plans sit in the center of a complex ecosystem, with various channels designed to communicate with each stakeholder group.
Health plan customers (patients) often need to reach out to their health insurance company (payer) for information regarding their plans. This specific information can be associated with their individual benefits, such as coverage eligibility, finding doctors, and claim reimbursement. Doctors and nurses also interface with payers for touchpoints like confirming coverage and negotiating payment through pre-authorization.
Payer investments catering to the needs of their constituent groups have yielded mixed results. Complex jargon/ terminology and the maintenance of archaic technologies and solutions (e.g. mainframe databases) lend themselves to the difficulty in creating a comprehensive solution. As a result, customers often have a subpar consumer experience that would not be tolerated in other industries.
Conversational technologies coupled with AI/cognitive computing technologies such as natural language processing/ understanding engines (NLP/NLU) and machine learning techniques offer an avenue to address this challenge. While they are not all new, these technologies provide innovative ways of navigating the complexities that have historically constrained traditional rules-based software systems.
Using AI training models, these systems can be continuously trained to answer new requests (queries) from customers and evolve into a robust Q&A system that is able to address the most frequently asked queries. Likewise, the response to an inquiry can be customized to match the level of detail and jargon appropriate to the consumer. For example, a plain English language response can be presented to a customer, while a doctor or nurse can receive a response that includes significantly more technical healthcare detail.
A typical customer query is, “what is the status of my claim?” Using NLP, a conversation platform can interpret
the request for claim status information and then subsequently retrieve and relay the related details for that particular customer and log the interaction for potential future use. This provides a much smoother customer interaction than calling into an interactive voice response unit that often fails to parse the customer’s utterances and only offers hardcoded options from which they can choose, which often results in customer frustration. Similar frustrations occur in call centers when customers call in about the same issue multiple times, and a lack of system coordination requires the customer to continually provide the same details about themselves and their situation to the call service representatives. Recent advancements in AI capabilities for dialog management using techniques such as Partially Observable Markov Decision Processes (POMDPs) improve significantly the ability to handle out context questions, to resolve ambiguity by predicting likely intent of human utterances, and actively learn from human interactions to improve accuracy and user experience. Some companies with large in-house development capabilities opt to custom develop those advanced features and lead the market in that respect, while others default to what is currently made available by the toolkit vendors and plan on incorporating those features as they are incorporated by the vendors. The obvious tradeoffs for these decision include complexity of use cases being tackled, size of investment and sophistication of technology and data science organizations.
From an investment perspective, such a platform can be built incrementally over a period of time to avoid the need for massive lump sum development investments. Engineering teams can add capabilities individually and incrementally, with a feedback loop and a continuous training process to enhance the solution as it matures.
Key Considerations for a Conversational Platform
Conversational platform consumer adoption is heavily dependent on response accuracy. A reliable and accurate response to the query is expected irrespective of complex back-end processing. Transitively, technological considerations are important when choosing a platform as the technologies are evolving with multiple vendors offering integrated solutions that offer different tools, features, and performance. A disconnect on such a paradigmatic platform with disregard for a unified protocol could pose a challenge for interoperability and data exchange. A similar challenge currently exists within the evolving blockchain space. Both technological ecosystems’ evolution will have an impact on the level of ease with which we are able to connect systems and transfer data, thereby influencing response accuracy.
It cannot be overemphasized how important it is to get right the User Experience (UX) of conversational systems. The shift from Graphical User Interfaces (GUIs) to Conversational User Interfaces (CUIs) is just as comprehensive and disruptive as the move from Web to Mobile and should not be taken lightly. The AI technology under the hood is the critical enabler to make it happen, but without thoughtful usability design and understanding the intricacies of conversational human interactions the systems are likely to fall short on adoption and business impact.
Organizational considerations within the conversational platform space include the use of closed-source or open-source platforms: for example, Rasa.ai is an open source NLP/NLU engine while Google’s DialogFlow and Facebook’s Wit. ai are closed-source platforms. Some industries may have little problem with third party platforms that collect metadata on the information that flows through their conversational engines given the robust nature of their solutions and feature sets. Industries like healthcare and finance deal with significantly more sensitive customer data that is required by law to be protected, and may need to build their own company/ industry standard based on open source platforms. Additional considerations exist with regard to application infrastructure for similar reasons; public cloud solutions such as Amazon’s AWS and Microsoft’s Azure are feature-rich and robust hosting solutions that offer similar data sensitivity challenges. Hosting solutions such as RedHat’s OpenShift and hybrid/private cloud solutions may offer improved data protection capabilities.
Conversational platforms in conjunction with other AI technologies offer opportunities to address the challenges in the Healthcare payer space that have persisted through multiple decades, problems that map to challenges that other industries experience. This technology is an important piece in building the future Healthcare ecosystem and may very well become a familiar part of our lives as consumers.