Main Library
In a fifteen-minute visit, there’s already a lot happening.
I’m listening to the patient, reviewing their history, asking questions, examining them, and trying to figure out what’s actually going on under the surface: the difference between what they’re telling me and what’s happening. Then I have to decide what needs to happen next.
That’s the part of medicine I trained for.
At the same time, I’m documenting everything in a way that has consequences far beyond that visit. Did I capture every relevant condition? Is the documentation specific enough? Does it support the diagnosis? Did I address the gaps that matter for risk adjustment and quality?
That work matters. I know why it matters.
But it’s not why the patient came to see me.
And that creates real tension for physicians. Every minute I spend working through documentation requirements during the visit is a minute of attention that has to come from somewhere else. Too often, it comes from the patient sitting in front of me.
Most physicians know exactly what that feels like. But more importantly, every patient knows what it feels like, too.
For years, the answer has mostly been to ask us to simply do more. Document more. Click more boxes. Respond to more queries. Keep track of more things that have been layered into the visit over time. Spend more “pajama time” at home in the evenings, time away from our families and hobbies — the lives we went into medicine to build and enjoy in the first place.
AI is supposed to make some of this easier. Sometimes it does. Sometimes it just gives me another list to deal with.
If a system surfaces twenty possible diagnoses during a visit, I still have to stop and evaluate twenty possible diagnoses. If it adds another window, another alert, or another workflow outside the EHR, then the burden hasn’t really gone away. It’s just moved, and it’s just taken more of my attention away from the patient in front of me.
The tools that actually help feel very different.
They bring forward the information I need when I need it. They show me why a condition is being suggested. They fit into the workflow and the EHR I’m already using. And they let me make a decision quickly without asking me to dig through years of chart history first.
Most importantly, the clinical decision stays with me.
AI can find patterns in the record. It can surface evidence I might otherwise have to hunt for. But I’m still the person who’s examined the patient, understands the context, and decides whether that diagnosis belongs in the chart.
That’s how I judge whether a solution is useful.
Not by how many things it can find. By whether it makes the work of caring for the patient easier instead of adding another task to the visit.
That’s also the question I think provider organizations should be asking as they evaluate AI: Does this actually reduce work for the physician, or does it just move the work somewhere else?
It’s not always easy to tell from a demo.
On August 27 at 1:00 p.m. ET, I’ll be talking about exactly that with Subbu Ramalingam, a healthcare expert who’s spent much of his career looking at these same issues from the payer and value-based care side. In our webinar, AI at the Point of Care: Best Practices for Evaluating AI in Value-Based Care, we’ll meet Margaret, a patient in my virtual exam room, and follow her through a typical visit. We’ll look at what I see as the physician, what happens after the encounter on the risk adjustment side, and what useful point-of-care AI looks like from both the clinician’s perspective and the payer’s perspective. Register Here for free.
If your organization is trying to separate AI that genuinely helps clinicians from AI that simply creates more noise, I hope you’ll join us.
Most physicians didn’t train to be coders. Good technology shouldn’t require us to become one.
Dr. Sunil Nihalani is a practicing physician and the founder and CEO of Inferscience.