AI chatbots might be the money-savers hospitals are looking for

But the technologies, which tap into natural language processing, knowledge management and sentiment analysis capabilities, require commitment and discipline to use well.
By Bill Siwicki
11:16 AM
Share
AI chatbots

Like other innovations, AI chatbots in healthcare will be a crawl-walk-run endeavor, where the easier tasks will move to chatbots while awaiting the tech to evolve for more complex work.

Healthcare provider organizations spend a lot of money on customer service representatives taking patient inquiries via phone, e-mail or live chat. But there’s a way technology can step in and save healthcare organizations time and money: automated chat-bots infused with artificial intelligence.

Among organizations in various industries, healthcare providers most of all will benefit from increased use of chatbots, which are becoming more adept at their work because of advances in AI, Juniper Research said. Chat-bots could save organizations $8 billion annually worldwide by 2022, up from $20 million this year, Juniper Research forecasted.

[Also: How AI is transforming healthcare and solving problems in 2017]

“We believe that healthcare and banking providers using bots can expect average time savings of just over four minutes per inquiry, equating to average cost savings in the range of $0.50-$0.70 per interaction,” said Lauren Foye, a Juniper Research analyst.

Most chatbots use multiple technologies: natural language processing, knowledge management and sentiment analysis. 

First, natural language processing tries to understand what a user is asking about. And second, a technological methodology provides conversational flow and responses, either direct or through guidance.

Typically, the natural language processing will identify the intent of a question with some level of confidence and then, based on the confidence level, the chatbot will either ask a follow-up or disambiguate the question for the user.

Once the confidence level is acceptable for the use-case, the chatbot will present the proper response based on an intent taxonomy that associates the intent of the question with the desired response. More advanced chatbots will try and anticipate the next question or guide the user to relevant resources or responses based on the previous intent.

“The technologies that support a chatbot need a common taxonomy in place that links the intent of a question to a contextual response,” said Jeff Cohen, co-founder and vice president of cognitive innovation services at Welltok, an AI-based healthcare software company. “And how do they interact to provide users with an answer to their question? There are many different ways to interact based on the sophistication and use-case for the chatbot.”

[Also: New partnerships are fast-tracking AI tech]

In addition to natural language processing technology, chatbots typically also rely on knowledge management systems.

“Knowledge management systems are absolutely essential in order to standardize the service experience,” said Khal Rai, an AI expert and senior vice president, product development and operations, at SRS Health, a healthcare software company. “Essentially, knowledge management systems are tools that allow you to document common questions and answers and problem-solving tips that are accumulated over the life of a product or a solution.”

It requires commitment and discipline by healthcare organizations to invest the necessary time and money to build knowledge libraries, Rai added.

Sentiment analysis is another technology that can be used by AI chatbots.

“How does the chatbot conjure up what is needed to be said?” asked Cohen. “Most AI chatbots need some content store or ‘traffic cop’ that knows, based on the intent of the question and the context of the user, where to obtain the proper response.”

AI chatbots have been used with varying levels of success in healthcare to date, addressing use-cases including helping consumers select a benefit plan, providing customer service responses, helping triage symptoms, and guiding consumers to resources. It still is early in the adoption of AI chatbots in healthcare, experts said, but early indicators of demand and satisfaction are promising.

“Chat-bots will continue to get more intelligent over time, thanks to AI and machine learning techniques that will make them very efficient technology, and of course, more timely than a human can ever be,” Rai said. “However, if you’re in the business of taking care of people, it’ll be a while before chatbots are fully adopted.”

Like other innovations, AI chatbots in healthcare will be a crawl-walk-run endeavor, where the easier tasks will move to chatbots while awaiting the technology to evolve enough to handle more complex tasks, Rai added.

“Research in the areas of emotional intelligence is happening,” he said. “But it is not advanced enough at this moment to put the satisfaction of customers on the line.”

Twitter: @SiwickiHealthIT
Email the writer: bill.siwicki@himssmedia.com