(left-to-right: Jason Peoples, Dr. Aaron Wilcox, Jonathan Friesen, Trace3Sara Sample-Reif

LAS VEGAS – We hear about AI all the time these days from vendors and other industry pundits singing its virtues while warning of potholes along the way. At Trace3 Evolve this week, practitioners told their AI tales.

The first day of Evolve Monday featured keynotes and panels focused on the healthcare and financial services industries. While these industries have different technology structures, the speakers shared common views about AI. The theme was it can be a great help in driving their businesses but requires a great deal of education – and perhaps patience – before it lives up to expectations.

Dr. Aaron Wilcox, a surgeon physician technology leader at the Southern California Physician Medical Group, outlined the challenges in finding the right AI solution.

“When you're being sold something, you think you're getting something like a Honda Accord, which is reliable, it’s there when you need it, it will go 300,000 miles, and you can give it to your kids when they turn 16,” he said during a New AI Tech Stack panel. “But that’s rarely the reality. Instead of showing up in a dealership and seeing a Honda Accord with a big ribbon on top, you probably see one of a few other things. You may see something that looks like a Fred Flintstone car. It looks like a car but a nurse or doctor has to do all the work, they have to be the engine. It doesn't have the power to transform. The second thing that you might see is the complete opposite – an F1 car. It's completely overpowered. It's not safe. It doesn't meet regular requirements. It hasn't been properly tested on the street. In other words, not street legal. You just can't use it.

“The third thing you may see is not a car at all, but you enter a shop and there's some windshields and tires and doors, and you have to build the car yourself. And so when we take this approach, what we really do is just break it down to the problem we're actually trying to solve.”

In other words, many AI solutions are a work in progress or wrong for your organization.

Evolve attendees heard success stories. Shawn Landreth, VP of network technology for Capital Group, described partnering with Trace3 consultants to build an AI-powered network dashboard on top of Splunk to better resolve connectivity issues. Dr. Adam Weinstein, CIMO of the DaVita kidney care company, talked about using AI to measure clinical ROI and to help shape human behavior for better patient outcomes.

But those and other speakers admitted AI is still evolving and still has challenges, particularly around data quality and regulations.

“I'm going to offer you unique perspective on how we are tackling AI,” Weinstein said. “What you're going to get today is a little bit of a vision of what it looks like in our AI oven. That is, we are far from fully baked. The lessons we've learned along the way are valuable to others.”

Data Quality Challenges

Weinstein said a big challenge for DaVita was “we have to have accessible and usable data. DaVita has about 30 years of dialysis data, which is many, many, many, many, many, many terabytes, and it is in varying states of usability. The capability to use that data is important and, of course, appropriate governance.”

On the financial services side, Ray Austad, head of operational risk management for Farmers Insurance, faces similar challenges around the amount and quality of data in his organization.

“We have 128 million records and it's shared between functions and business units,” he said. “From a data quality perspective, we’re not there. When we start to accelerate into AI, what are we going to get? We expect accuracy and timeliness of the data. That's one of the things that we're struggling with right now. I feel we're being very tactical in our approach, but at least we're focused on it, which is great. Data quality is tied to a lot of things, including bias.”

Capital Group’s Landreth said because of his company’s variety of data sources, there were no commercially available dashboards he could use.

“Nobody had a dashboard that had everything we wanted to see,” he said. “Looking at a variety of data sources was our challenge. How do we bring all these into one view? The data quality piece will influence what comes out on other side.”

Data Governance: AI Isn’t Magic

While most of the practitioners mentioned governance as an issue, a lawyer on the healthcare panel lives it every day. Jonathan Friesen, associate chief legal officer and chief privacy officer of Geisinger Health, said compliance teams should be involved early in all AI projects.

“I think it's an extraordinary time to be working in this space,” he said. “I think it's going to revolutionize healthcare in this country and in the world. Obviously, there are a ton of issues from a governance and compliance standpoint that keep me up at night. We want transparency. We want to know how these things work. We want to know what data does a system have access to, how is it using the data? Is it transforming the data? How is it going to affect our privacy and security, compliance and protect everyone's data?

“A lot of people think that AI is magic and goes into a processing center and spits out a result, and they can't tell us what's actually happening. That is a huge red flag from a legal and compliance standpoint, because we have to be able to explain it [to regulators].”

Like many in the tech industry, the Evolve speakers expect challenges will be overcome and AI will be a boon to their organizations.

“We're in the business of fixing problems, not implementing solutions or implementing technology alone,” Wilcox said. “So we try to root things in reality. We try to root them in the human experience and build things that not only fix a problem for today but can fix a problem in perpetuity.”