Last December I spent a full day on-site at a health clinic doing workflow interviews. I had a Plaud recording pin on the whole time — the goal was to stay present in every conversation while still capturing everything.
One of the interviews I did that day was with the clinic's head physician. She runs a boutique concierge medicine practice, sees premium patients, and manages a small clinical team. I asked her about clinical decision support tools — whether she used any, whether she wanted any.
What she told me stuck with me.
She's already using it
She uses Gemini and ChatGPT to write patient counseling summaries. Instead of spending 20 minutes per patient developing her own written narrative — lifestyle advice, dietary notes, whatever the patient needs to hear — she asks AI to draft it in five sentences. Then she reviews, adjusts if needed, and shares with the patient.
She also uses the AI scribe built into her EMR. It transcribes consultations in the background while she talks with patients. Before she had this, she was charting during appointments — trying to type notes while simultaneously listening and responding. Not ideal for anyone in the room. Now she can be fully present.
And she uses AI to summarize lab trends across a patient's history. Her EMR has a clinical assistant that can pull up changes in liver enzymes or blood panels over time. Instead of manually scanning through hundreds of reports to find the relevant patterns.
None of this involves diagnosis. All of it is paperwork.
The line that changed how I think about this
“My time should be used for diagnosing,” she told me. “Not charting.”
That one line names something I've been watching play out across dozens of client engagements. AI creates the most value in professional work not by replacing expert judgment — the part that requires training, experience, and accountability — but by eliminating the overhead that surrounds it.
Every knowledge worker has this split. There's the expertise work, the stuff that actually requires you. And then there's the overhead: writing things up, organizing information, following up, summarizing, explaining, preparing. That second category is where a lot of the hours go.
For a doctor, the expertise work is diagnosis. The clinical judgment that requires years of medical school and a license. Everything else — the charting, the write-ups, the record review — is overhead.
The same pattern shows up in every profession. A lawyer's expertise is legal judgment. The research, the drafting, the summarization is overhead. An executive's expertise is strategic decisions and key relationships. Calendar management, email triage, meeting prep is overhead.
AI is best at the overhead.
I've seen this across a lot of different clients
Last year I was working with an executive team on AI automation. One of their assistants was spending 20-25 hours a week on administrative tasks — email, scheduling, meeting follow-ups, basic research. We built a set of AI agents to handle the routine work. Within a few weeks, that same work took 4-5 hours instead.
She didn't lose her job. She got her time back for the work that actually required her.
That's the version of this story that plays out when people use AI well. Not handing off the judgment calls. Handing off everything that was never really a judgment call in the first place.
The question most people ask about AI — “will it replace me?” — usually points to the expertise work. Will AI diagnose patients? Will it argue cases? Will it make strategic decisions?
In most fields, for now, no. But that's not where the time is going anyway.
The time goes to the overhead.
A useful test for your own work
Before your next task, ask: does this require my expertise, or just my time?
Writing up meeting notes after a call — that's time, not expertise. AI can do it.
Deciding what the next move should be based on those notes — that's expertise.
Drafting a patient counseling summary — time, not expertise. The doctor already knows what the patient needs. She just needs it written down.
Knowing what the patient actually needs — expertise.
That split is different for every person and every profession. The doctor figured out where her line was. A doctor with a different specialty might draw it differently.
The professionals getting the most out of AI right now aren't asking whether AI can replace them. They're asking which 20-minute tasks they've been doing that never required their expertise in the first place.
That's the question worth sitting with.
Want to map out where the overhead is in your own work? The Two Hour Workday program walks through exactly this — identifying which parts of your day require you versus which parts AI can handle. More info here.
