.svg)
I’ve spent nearly 30 years in technology and medical imaging. Most of that time, I’ve been the one introducing the technology. A better system. A better workflow. More information. More capability. More power for clinicians to do what they do.
And for almost as long, I’ve been puzzled by resistance. Why wouldn’t someone want a better tool? Why wouldn’t they embrace technology that could make their work easier or give them capabilities they didn’t have before?
After a recent ride in a Waymo, I think I understand that resistance a little better.
This year, I attended my first College of American Pathologists annual meeting as CEO of HyperPath. Digital Pathology and AI were, not surprisingly, a prevalent topic . But the most interesting conversations weren’t really about what AI can do. They were about the open questions.
Where should it fit? What should pathologists embrace? What deserves skepticism? What foundation needs to be in place before AI becomes part of routine practice?
And underneath all of that, how do we introduce increasingly powerful technology without taking control away from the people responsible for patient care?
When the ride changed
A few hours later, that question became very real to me. My colleague Shanda and I needed to get from the hotel to the conference, and we decided to try Waymo. Why not?
The experience started beautifully. The car arrived. We got in. We selected our destination. And the car drove. It was genuinely impressive. Smooth. Confident. A little surreal. For someone who has spent his career around technology, it was hard not to enjoy the experience.
Then we got close to our destination. Construction and traffic rerouting around the conference area appeared to confuse the system. We were almost exactly where we needed to be. And the car missed the turn.
Fine. Technology makes mistakes.
The vehicle recalculated the route and sent us on roughly a 20-minute loop through the city so it could approach the destination again. So around we went.
Twenty minutes later, we came back. And it missed the same turn again.
Then things got more interesting. At one point, the Waymo reached an intersection with a green light. And stopped. It put on its flashers.
Cars began stacking up behind us. People started honking. And the car just sat there. It appeared to have no idea what to do next.
That was the moment the experience changed for me. A few minutes earlier, I was marveling at an autonomous vehicle. Now I was trapped inside one.
There was no driver to tell, “Forget the route. Just pull over here.” At one point, even the option that normally allowed us to request that the vehicle pull over was no longer available.
The system was making the decisions. I didn’t know what it was trying to do. I couldn’t change its plan. And I couldn’t simply tell the driver to stop, because there wasn’t one.
We called Waymo support. Eventually, the vehicle started moving again. It continued its route, brought us back toward our destination and essentially repeated the same problem. By this point, it was preparing to send us on yet another long loop through the city.
We called support again. Eventually, human intervention helped get the vehicle to another drop-off location where we could finally get out and walk the rest of the way to the conference.
Nothing terrible happened. Nobody was hurt. And I would absolutely ride in a Waymo again. But my perspective on the technology had changed.
For the first few minutes, I was watching an impressive autonomous system. Later, I was dependent on it. There is a very big difference between those two experiences.
The pathologist is inside the car
And somewhere during that ride, I had a somewhat uncomfortable thought. This may be exactly how physicians sometimes feel when people like me show up with new technology.
For decades, we’ve walked into hospitals excited about what we’ve built. Look at this. Look what this system can do. Look how much better this could be. And most of the time, we mean it.
But there is an enormous difference between demonstrating technology and being dependent on it.
The pathologist isn’t watching our demo. They’re inside the car.
A real patient is attached to that workflow. Their judgment matters. Their name goes on the diagnosis. And when the technology encounters something its designers didn’t anticipate, they still own what happens next.
That changes the standard. A system that performs beautifully most of the time sounds impressive in a presentation. It feels very different when you are personally sitting inside the exception.
Control is a trust question
We spend a lot of time in healthcare asking whether technology works. Maybe we need to spend more time asking whether the person using it still feels in control.
- Can they understand what the system is doing?
- Can they challenge it?
- Can they override it?
- Can they stop it?
- Can they recover quickly when something goes wrong?
- Does the system recognize when it is repeating the same mistake?
- And when automation reaches its limit, does it know when to hand control back to a human?
Those aren’t soft questions. They are trust questions. And trust may ultimately determine how quickly digital pathology and AI move from impressive demonstrations into routine pathology practice.
There is a temptation in technology to interpret resistance as fear of change. I think that is often too simplistic. Sometimes resistance is completely rational.
A pathologist may not be protecting an old workflow because it is familiar. They may be protecting their ability to exercise judgment. Their ability to intervene. Ultimately, they may be protecting the patient.
Successful adoption of AI in pathology therefore can’t simply be about creating better algorithms. It requires the right digital foundation. It requires interoperability. Governance. Thoughtful exception handling. And systems designed to keep the physician in control rather than turning them into a passenger.
The best technology should give clinicians more control, not less.
When automation reaches its limit
There was another lesson from that intersection too. When an autonomous vehicle stops at a green light, turns on its flashers and appears completely confused while traffic piles up behind it, the quality of the technology suddenly matters a little less. What matters next is support.
Who answers? Do they understand what is happening? Can they intervene? Can they actually get you out of trouble? Can they solve the problem quick enough to not make matters worse?
That experience changed the way I think about what genuine customer support needs to look like when customers depend on sophisticated technology to do critical work. That’s a story for another post.
And while we’re on the subject, a small piece of unsolicited advice for anyone who encounters a driverless car stopped in an intersection:
Honking at it doesn’t work. Rolling down your window and yelling at it doesn’t work. And, based on what I observed, the traditional one-finger salute is equally ineffective when there is nobody behind the wheel. Save the energy.
The lesson for AI in digital pathology
For this post, though, my lesson was simpler. After almost 30 years of bringing technology into healthcare, a driverless car gave me a much better understanding of what it feels like to be on the other side of a technology pitch.
If we’re going to ask pathologists to trust digital pathology, automation and artificial intelligence, showing them that the technology can drive isn’t enough. We also have to make sure they know they can take control when the road doesn’t go according to plan.
The next ride
And apparently I’m not quite finished testing the theory.
I’ll be heading to San Diego next week for the Digital Pathology Association’s Pathology Visions conference, October 16–18. Waymo recently began offering rides to members of the public there as it expands its San Diego service.
So, assuming I can get access, I’m going to do the completely rational thing. I’m taking another Waymo.
Stay tuned.