Meet AI-Powered Olive, a Digital Employee Who Wants to Take on Repetitive Tasks in Healthcare

Olive is a healthcare bot that’s built not just to work with humans, but to seem like one. The company describes “her” as an AI-powered, digital employee to automate repetitive, high-volume, error-prone workflows.

Olive’s CEO Sean Lane developed software for national security before switching to healthcare. Olive is Sean’s solution to the interoperability problem in healthcare.

Here are the edited highlights from my podcast interview with Sean about Olive’s origins, how she can fight the opioid epidemic and why we shouldn’t be afraid that she’ll steal our jobs.

From Ohio to Afghanistan and back

Olive is AI for healthcare, but you come from a national security background. How did you get from there to here?

I was in the intelligence community. I spent most of my career at the National Security Agency (NSA), which was a longtime dream for me. I spent five tours in Iraq and Afghanistan building pretty large systems.

I got out of the NSA in 2007. I started a company building software for the intelligence community, and I sold that company around 2012. I was looking for what to do next.

At about that time, my hometown in southeastern Ohio — the small town of Gallipolis — had a big problem with prescription drug abuse. They were looking for solutions and they knew I had a tech background. They asked if there was some technology that could help.

When I looked at healthcare information systems, specifically the hospital in my hometown, I noticed that it was kind of like the intelligence community before 9/11: no sharing of information, its systems didn’t talk to each other, no interoperability.

Systematically, that led to many of the issues that we were having, not just with prescription drug abuse and diversion, but across the board in healthcare. I made it my mission in life to figure out how to connect those things.

I decided that the best approach was to create a software robot using computer vision — making deep learning that interoperability layer, so the robot can log in to any piece of software like a human does. It can operate it, connect the dots between various pieces.

Can you be more specific about how data challenges and the lack of system connectivity leads to the prescription drug problem?

In healthcare, big enterprise systems are like fortresses. Data sits inside those fortresses. Over time, the healthcare industry has acquired lots of different software systems. But the databases between those systems don’t commingle. So the ability to do entity resolution at scale, to determine that you’re talking about the same patient taking the same drugs, with the same diagnosis, is extraordinarily difficult.

With a software robot, we can log into systems just like a human would, and then do entity extraction and entity resolution, connecting the data together across different software systems.

Turing-tested, human-approved

You talk about Olive in human-like terms — as a digital employee. You even give Olive the female gender. Why did you decide to do that? How much of it is about the way that the system actually works? How much of it is marketing?

To solve the interoperability problem, we spent all this time building really sophisticated technology. Then we had to put it into practice.

In the past, developers would do that by creating software, which in turn created another user interface. People in healthcare already have 20 — maybe 50 — user interfaces to do their job. The last thing we wanted to do is introduce a new one.

So how do we manifest this technology into practice? The idea was that we really need to take over some of the administrative tasks that humans are doing. Healthcare’s plagued with what we call ‘human routers’: people doing repetitive, mundane tasks over and over again.

We thought, what if we were to create a digital employee to take over a lot of these jobs? One who can enter the HR system, get credentials, get an email address at the organization, just like a human employee.

But we basically armed that digital employee with entity resolution and computer vision capabilities, so it’s able to see a screen and understand what’s on it. Then we added a whole host of data feeds, dictionaries, ontologies and deep-learning algorithms, which made it a super employee that could do anything. We were riding the edge of artificial intelligence.

I’m kind of an old-school AI person, in that I subscribe to the Turing AI methodology. That means it should be very difficult to distinguish the artificial intelligence technology from a human. That became one of our core focus areas: if we’re going to do this successfully, it should pass the Turing test. You shouldn’t be able to distinguish our software robot from a human.

That’s why we gave it a persona. Olive works at your organization. She reports to a manager. She shows up for work every day. She does her job extremely well. She gets smarter over time.

When [IBM’s] Watson won the game of Jeopardy, I remember thinking: well, that’s not really a person. They gave it a name and a voice as a nice marketing tool. But in so doing, it’s not just marketing. It makes people accept and understand how to work with that technology in a different way, which is just as important, if not more so, than the technology itself — getting people to actually adopt it.

Are you touching on something that goes beyond the healthcare space? Are you demonstrating how to use AI in ways we haven’t recognized yet?

I would say that we’re showing a deep vertical version of a software robot. We certainly didn’t invent the idea of software robots. Over the past few years, a technology called RPA, robotic process automation, has become incredibly popular and distributed at an enterprise level across many industries. Right now, though, it’s a tool you use to create a robot that does a task, but it doesn’t quite take you to the level of a digital employee.

We’re using the underlying foundational technology of RPA and take it really deep and really verticalized in healthcare. For us, it’s important to focus on a particular industry so that we can build a digital employee that goes well beyond automation.

Robo-boss vs. ‘scaling humans’

You mentioned that Olive reports to a manager. How long do you think it’ll be until Olive is managing the hospital?

I think that Olive will continue to take on more responsibility inside an organization.

Olive is part of an approach we call ‘scaling humans’.

We haven’t seen the capabilities of human workers yet because haven’t had the opportunity to scale them by taking away their [mundane], robotic tasks. I don’t think Olive will ever be at the top of the food chain. Humans will always be there. But I think she will be able to take on more responsibilities as time goes along.

I’m thinking about the classic question: “Are robots going to take my job?” Do the people you sell to think of Olive as a threat to human jobs?

Oh, yeah. It’s a conversation we have every single time, with every customer. Because once you see Olive working and really realize her potential, your mind immediately goes to the inevitability that Olive will be doing a lot of the work that humans are doing today. It really takes a reframing of how we think about work in general.

Right now, just figuratively, there are 100 things the healthcare industry needs to accomplish. Right now, it’s only getting to 50 of those. There’s a lot of capacity that still needs to be filled. That’s where digital workers come in.

But over time, we’ll have to start up-skilling human employees to take on new, different tasks. Digital employees won’t be taking away their careers, but they will be taking the jobs that humans are doing today and humans will shift to work on something else. That’s a good thing. That’s what we’ve always done as a human race. That’s what we’ve done as an economy, as a society. We’ve always shifted work. The digital worker is an important part of doing that right now.

Don’t miss the next podcast episode! Get email updates.

VP of Marketing & Communications for Infinia ML, a machine learning company. Speaker from North Carolina to South Korea.