A number of chatbot offerings were demoed at a special AI-focused session of the Health 2.0 annual conference in Santa Clara, California on Monday. Although each had a unique take on how best to insert smart conversations into patient care, the growing number of these technologies suggests a clear trend, panel moderator and Chief Product Officer at Seniorlink George Kassabgi suggested.

“We’re in the middle of a new paradigm shift for interfaces,” he said during the session. “Rather than installing an app, filling out forms, working with menus, you’re simply typing or talking with a system. That’s a pronounced shift, a really meaningful change.”

AI has been deployed in a greater and greater number of use cases in digital healthcare, and according a recent analysis will see as much as a 40 percent annual growth rate between now and 2021. Interest in leveraging this technology to improve messaging and communications is increasingly growing, Kassabgi said, and well-designed conversational interfaces are poised to change the game.

“None of these [panel participants] are demoing an app, per se,” Kassabgi said. “What they’re going to be demoing are conversations. If you take one thing away from this session, it’s that conversations are the new thing.”

The companies participating in the session included: Buoy Health, Catalia Health, Memora Health, SapientX, and SimplifiMed. Each gave a short demonstration their AI chatbot-based technologies, which ranged from SimplifiMed’s mobile SMS-driven self-diagnosis and referral system to Sapient X’s voice chat triage platform, that users see as a computerized face on their computer screen.

But not all bots are created equal. While there have been a number of chatbot-based applications over the years, many have been unsophisticated tools weighing basic mathematical probabilities against datasets, Sapient X’s CEO David Colleen lamented.

“We’ve seen in the past few years with AI in the news headlines all the time, there’s so much hype out there, chatbot companies being sprung up left and right,” Colleen said during the panel. “Most of them are using technology that’s from 1965 … they’re pretty primitive.”

Even the next generation of neural network-based systems has its flaws as well, he continued. Although these systems are very capable of word matching, they consistently fall short when it comes to conversations. The chatbot system of the future, he said, will need to avoid command structures and build systems that mirror how the human mind thinks contextually, and will likely be some kind of hybrid of past strategies.

In the following discussions, the panelists also agreed that widespread implementation of chatbot-based systems could improve data collection. Clinical practices could be honed based on longitudinal data on patients’ health markers and daily habits collected through chat logs, and the candid conversations users have with bots could allow the conversations to be examined for new approaches in care.

“Something that we think is very important is the metadata around the conversation,” Nisarg Patel, CEO of Memora Health, said during the session. “We think it’s important to see what types of conversations and triggers can really influence patients to change their behavior, and what really gets them to open up to clinicians about what they’re suffering through. We find the timing and content, as well as the kind of interaction they have, is not only important in guiding what we’re building but also how clinicians should make decisions about care.”