LifeLink CEO Greg Johnsen: AI Guts the ‘Fat Belly’ of Healthcare Costs

Original article was published by James Kotecki on Artificial Intelligence on Medium


Audio + Transcript

Greg Johnsen:
It’s all about using technology to free smart minds up to do the things that they’re really good at.

James Kotecki:
This is Machine Meets World. Infinia ML’s ongoing conversation about artificial intelligence. I am James Kotecki joined by the CEO of LifeLink Greg Johnsen. Welcome Greg.

Greg Johnsen:
Thank you. Good to be here James.

James Kotecki:
So Greg, your website says “chatbots are the future of healthcare.” So we’re talking about AI powered chatbots that don’t take the place of doctors it sounds like in your case at least, they take the place of an administrative conversation that I might have with somebody about billing for example, is that right?

Greg Johnsen:
That’s right. We’re not using our technology to replace a doctor’s diagnostic assessment or the skills of a physician. What we’re really focused on are all the workflows and protocols that are robotic that humans have to do. I’m talking about things like helping patients complete forms, reminding them to get to visits, doing intake, and triage, and questionnaires, and surveys and keeping them posted on their status, being a search engine for them so when they’re looking for something. That’s what we call the fat belly of cost and aggravation in a lot of the healthcare business.

James Kotecki:
Let’s go right to the million dollar jobs question that I’m sure you get all the time, “Is my job now on the line? I’m calling people all day trying to track down bills or set up appointments.” What do you tell those folks?

Greg Johnsen:
Well, it’s not like that work is pleasurable work or the best way to take a human brain and deploy it there. And what humans are really good at is all the empathy, empathetic discernment that has to happen in human dialogue and conversations and interaction. So if you can offload a lot of the rote procedural work that burns them out and move it to a mode that is low cost but also one that patients prefer. So, the way we explain it and articulate to large healthcare systems and pharma companies and healthcare organizations who we target is, “We’re going to shift up the skill level. We’re going to take the human teams that you have and allow them to shift up into the upper cognitive stack of what humans are really good at. And we’re going to take out the robotic work and move it to this layer that is not only cost effective but it’s the kind of engagement that consumers prefer to do with a digital assistant.” So this is about how to increase capacity in an industry that has been short and thin on service and capacity and how to solve that problem.

James Kotecki:
And we’re talking for context about people interacting with these bots via their keyboards and screens, right? We’re not talking about an automated voice calling them on the phone or are we?

Greg Johnsen:
That’s right. We’re predominantly, exclusively what we’re doing right now is chatbots in chat on mobile phones. You’re going back and forth with chat bubbles.

James Kotecki:
So, you’re saying that this is work that humans generally don’t want to do or aren’t best utilized for and I tend to agree with you there and yet it’s interesting that the interface that you’re building is meant to in effect replicate that interaction people have with a human. Because if I go back and look at technology for the last 20 years, we’ve long had the ability to fill out a form on a website, right? But why are you choosing to do it in a human-like way? And as a follow-up question, how human are you trying to make this thing seem?

Greg Johnsen:
When you give a consumer, getting ready for a healthcare appointment, a complicated portal, or a set of steps to log in and download an app and learn it and find their username and password and register — it doesn’t happen. It’s why healthcare systems have about a two percent penetration in terms of reaching patients in workflows on mobile phones. Two percent. And it’s because it’s hard. Conversations are simple and they’re easy and they’re natural and there’s nothing to learn. It’s a powerful modality to connect with consumers at scale. And so, our whole architecture is built around that very simple idea which is how do you connect with a patient, with a consumer in a complicated healthcare flow with a conversation that feels like a human conversation?

Greg Johnsen:
Now to be clear James, we’re not setting up these chatbot workflows to fool anybody. It’s not like the entity on the other end of that conversation ever represents itself as anything other than a digital assistant. They feel like the thing you’re working with is doing things for you and can kind of anticipate what your next move is or what you might be interested in doing or what you should do. When we start talking about AI and machine learning this is really where LifeLink is focused. It’s using smarts and technology to predict what the next best thing or move is in the context of a specific healthcare protocol and a specific human, a person, a patient who we know about.

James Kotecki:
Are we really talking about two separate AI systems here? One system is the one that can go into healthcare records or billing records and find the information that is relevant for that next moment in the conversation, maybe it’s being queried specifically like a search engine. And then the other kind of AI, which is being merged together in your product, is that conversational part that makes it seem more natural and more like a conversational interface. Am I right to think of those as two separate things, and if so which one is harder?

Greg Johnsen:
Well, they’re certainly different applications of AI. The front end thing you might see or associate with AI with chatbots is natural language processing which is understanding what the patient — what the consumer is wanting to do. And so, you have to use AI to understand what their intent is, their intent and in the context of the entities or things that this domain is all about.

Greg Johnsen:
There’s also this idea of how to make smart predictions about where things can go. So for example, one of the big things in healthcare is patients not showing up for surgery, expensive. There’s an OR that costs $100 a minute. If you can get patients to show up on time and be ready it’s a big deal, but you need lots of humans to do that, to call people, to give them their instructions, to remind them, to call them every day, to get a ride for them. It’s raining, make sure you get a ride. That’s complicated stuff but if it’s chatbot-driven then it can happen at scale. And the AI used to make that happen is all about predicting which patients are most likely to not show up or to show up late or to show up unprepared.

Greg Johnsen:
So if you can begin to have different conversations with patients based on what you know about those kinds of patients and these kinds of workflows at scale, suddenly those conversations get a lot more powerful because they’re not just natural, well-timed conversations. They’re pinpointed, precision conversations that tie and stitch back to a value proposition. Fewer no-shows. Fewer late-shows. For example.

Greg Johnsen:
There’s two different applications of AI and there are 17 more. I mean, predicting wait times in an emergency department is another application. It’s another way of — it’s in that camp of machine learning where you’re looking at all the tests and the turnaround times for lab tests and blood tests and imaging tests over millions of patient visits, which is what we do. So you can communicate that to the patient and say, “Looks like your lab test just went in. It looks like it’s going to be about a 35-minute wait. You might want to get comfortable and we’ll let you know when you’re about five minutes away from getting your physician to come out and talk to you about it.” That’s a lot of computing to make that happen.

Greg Johnsen:
One of the really fascinating things about this kind of technology in healthcare is that it’s such a laboratory for human decisioning, human bias, human behaviors. How many nudges it takes to reach out to somebody to get them to engage, and does that have anything to do with their zip code? Does it have anything to do with their gender, or their age? You’re digitizing all the conversations and all the sentiments all along these workflows, so suddenly you have this consumer research platform. Are there patterns about, should we be incenting certain consumers in certain areas to get a wellness visit more often and should we pay for their ride?

James Kotecki:
Do you run into any issues of AI bias?

Greg Johnsen:
Probably less than in other AI areas. We’re not generating language. We use NLP to understand and then using predefined responses. And by the way, they have to be in healthcare. We’re literally tuning into the approved, medically compliant workloads and dialogues that have to be delivered.

James Kotecki:
One last rapid question before you go, how far along the path do you think we are to artificial general intelligence based on the work that you’re doing now or is that not even really even the right question to ask?

Greg Johnsen:
Clinical care is better than it has ever been in healthcare and we have highly-trained physicians who know more, have access to more than ever before. The problem is they’re not freed up to do that work. They’re encumbered by a lot of the administrivia and inefficiency of workflow. So, I really think for the next decade it’s all about using technology to free smart minds up to do the things that they’re really good at. And I don’t have an opinion on when general intelligence will arrive, I would only be speculating.

James Kotecki:
Well, that’s what we love to do sometimes but hey Greg Johnsen, CEO of LifeLink, I really appreciate you joining us today and sharing your insights here on Machine Meets World.

Greg Johnsen:
Thank you for having me. Nice meeting you James.

James Kotecki:
Thank you so much for watching. I’m James Kotecki. You can email the show mmw@infiniaml.com. Like us, share us, comment, you know what to do. That’s been what happens when Machine Meets World.