We’re in a moment right now, a moment of what investors call FUD (fear, uncertainty and doubt) around AI. Or maybe it’s more like fear and loathing. It’s something a bit more fundamental than investment jitters. Whether it’s a boom or a bubble, humans are becoming worried that AI will get too good at what it does, and cause any number of imbalances for which we are just not prepared.

But let’s take a look at the positive side, in actual improvements coming to the world of healthcare, where AI has been so instrumental in saving lives. Not just a few lives. In nearly all aspects of a complex system, AI has been helping clinicians to move the ball forward, treat patients with more care, and universalize the advantages of having Western medicine at your fingertips.

Thoughts from Silicon Valley

At the Next Endeavor event Sept. 14-15, in Mountain View, California, which Imagination in Action and partners put on (disclaimer: I help to run these events), Caty Byerly Rezendes interviewed Hippocratic AI cofounder and CEO Munjal Shah about some of these brilliant move-markers that we’ve seen in the healthcare sector. But the two started out talking about safety.

“There’s a lot of talk around AI right now, and safety, and you think of medical and healthcare as, you know, the sacred space,” Rezendes said. “So how do you actually build AI that’s safe enough for the healthcare space?”

In response, Shah told the story: of hiring 7,000 US licensed clinicians to look after the AI’s work in a telemedicine trial.

“(They were asked) to call the AI, act as a patient, mark every error it made,” he said. “One out of 50 times, they got a human nurse on the other line. and they marked the errors they made. We did not ship till we were safer than a human nurse.”

The whole thing, he said, involved some 280 million patient interactions.

Shah also told another story, of getting buy-in for this abundantly safe system.

“We went to the investors when we raised the money, and said, ‘We’re not going to ship till we’re safer than a human,’” he explained, “and they’re like, ‘Well, how long is that going to take?’ And I was like, ‘I don’t know. I’ll let you know when I get there. If you don’t want it, don’t give me your money.’ And we did that upfront before they committed.”

So what has AI been up to in this space? Here’s what I got from the presentation:

Pre-operative Instructions or Post-discharge Follow-up

When a patient needs help understanding what to do before a procedure, AI can help bring this assistance to the point of delivery. Same with home care afterward. If a human isn’t available to cover all of this, AI provides the granular instruction that can lead to better outcomes, shorter recovery times, etc.

So many drugs and so many rules: and again, if a doctor or nurse can’t be there to explain it all to you every day, AI concierge services can.

AI can also be a great way to accomplish those test protocols that represent a high quality of care in a clinical environment. Shah explained:

“We actually run a protocol called the Schmidt Thompson protocol, which is - 95% of nurse triage call centers in the U.S. run that protocol. If something happens, you say, well, I ran Schmidt Thompson. So that’s a bit of the standard of care that gives you a little bit of liability protection for a health system.”

Rezendes noted that this kind of checking has been part of traditional medicine.

“Sometimes it’s very simple,” she said. “I recall surgical checklists that came out of Johns Hopkins, and it’s just making sure that you’re performing surgery on the correct limb, or actually marking that it’s the right limb, not the left limb … there’s very analog redundancies built into the system, which is reassuring to some degree that that’s built in, but it also seems like it’s something that’s really ripe for reinvention.”

Attracting the High Risk Patient

Here’s another one: how do you get high risk patients to come in and get checked for those important status updates? Well, it turns out AI can be great at that too.

“The AI was so persuasive,” Shah said, in an enumerative summary. “It got 252 people to do it, to schedule. 180 showed up. Nine significant findings. Three cancers. … the AI has empathy at all times and never loses its cool.”

That’s five solutions, if you count the pre-op and post-op stuff separately.

“The rising tide is just lifting all boats in AI,” Shah said. “(But) we’re just so used to zero sum thinking. We don’t naturally think this way … but what happens when everybody (has) something looking out for you, every day? Trying to make you healthier?”

What happens, indeed? We’re finding out. Stay tuned for more.