The Human Touch in an AI World: Why I'm More Excited Than Scared
Over dinner last week, I was telling my wife about a patient I’d taken to the OR for a uterine biopsy. It’s a procedure I don’t do as often as I did during residency, so I didn’t have my standard postoperative instructions ready.
“I used OpenEvidence to generate evidence-based discharge instructions,” I told her. “Took less than two minutes. Between that, Doximity’s AI tools, and DAX for ambient scribing, there’s no telling how much time I save in a day compared to a year ago.”
She nodded and told me about using DAX in her reproductive endocrinology practice. Then we got onto what’s coming next. Ambient ordering within a couple of months. Whatever follows that.
We’re both physicians watching this change our daily work in real time. I’m thrilled, and some nights I lie awake wondering what it means for the profession our son might grow up around.
What the tools already do
The diagnostic algorithms catch things I might miss. The documentation that used to eat an hour of my evening now takes minutes. I’m not speculating about this; I’m describing my Tuesday.
That same speed changes what a physician is for. A lot of what took me years to learn, a model now does in a second. So I keep asking myself which parts of this job a machine can’t touch, and I want to spend more of my day there.
Holding someone’s hand
OpenEvidence can suggest a treatment. It can’t stand in the operating room and hold someone’s hand while they fall asleep for surgery. It can’t look at a frightened patient and say, “I’m glad you’re here. I think we can fix this.”
So I’m trying to give that part more room.
With my older patients especially, I slow down and ask who they are. Where they’ve been, what they’re proud of, what scares them. I hear the same kind of story all the time, where a patient has quietly given up something she loves because of a problem she assumed she had to live with. Most recently it was a woman who’d stopped going fishing, something she’d done her whole life, because she couldn’t get through a morning on the water without leaking. That changed what “better” meant for her, and it changed what I offered.
The treatments I do are rarely one-size-fits-all. What a patient picks has to fit her values, her appetite for risk, the shape of her life. I can’t sort that out from a chart. It takes a real conversation, and the AI handling my notes is what buys me the time for one.
I perform surgery every day, so it’s easy to forget that for the woman on the table, this is one of the biggest days of her year. She deserves a doctor who remembers that, not one half-watching a screen.
And there’s a kind of judgment I don’t know how to hand off. The sense that something’s wrong before the numbers say so. Reading what a patient won’t say out loud. Knowing when the standard protocol doesn’t fit the person in front of me. I built that over years of being in the room, and I don’t think it lives anywhere a model can reach.
Where I’m pointed
As a urogynecologist, I’m not after a lifelong relationship with my patients. I want them better and free to forget me, except maybe when they’re telling a friend that treatment gave them their life back. Getting there still takes everything that makes us human: listening, understanding, caring about one person’s story.
AI won’t replace what passes between two people when one trusts the other with her body and her fear. I think it gives me more time for exactly that.
That’s why I’m more excited than scared.
How are you thinking about AI in your practice? I’d love to hear from other physicians working through it.