Before the first patient

There is a particular optimism in an empty clinic. The paint is fresh, the chairs are still wearing their plastic, the schedule is a clean grid waiting to be filled, and the founder, almost always a physician who has spent the better part of a decade wanting exactly this, walks the rooms with the look of someone who has finally been handed the keys to their own judgment. I love that moment. I also know how much of it is about to be spent badly, because the empty clinic is the one place in healthcare where the future is still genuinely open, and almost everyone fills it the same way the last clinic did.

The instinct, when you have nothing yet, is to copy. You hire a front desk because clinics have front desks. You buy whatever record system the consultant put in front of you, sign up for the cheapest phone line that will route a call, and tape together the rest as patients start to arrive. None of this feels like a decision. It feels like setup, like the boring necessary scaffolding you have to get through before the real work of medicine can begin. And so within a few months the founder who swore they would never run their practice the way the last hospital ran has quietly rebuilt the same administrative organism they fled, patch by patch, without ever choosing to. The clinic was empty for a moment, and then it was full of other people's habits.

We meet most practices long after that moment has passed. By the time an owner sits across from us, the phone has been ringing for two years, the voicemail box fills faster than anyone can empty it, and there is a fax machine in the corner that someone is genuinely afraid to unplug. They want us to point at the loudest problem and put it out, and the work of those first conversations is usually to slow them down, because the loudest problem and the most automatable problem are rarely the same thing. That is its own subject. But every so often the conversation goes differently. Every so often someone calls us before they have seen a single patient, before the chairs are unwrapped, and asks not how to fix the fire but how to build so the fire never starts. Those conversations took me a while to understand, and they have ended up teaching me more about what we actually do than years of firefighting ever did.

Here is the thing that took the longest to see. When you automate inside an established clinic, you are almost never adding something clean to an empty space. You are reaching into a living system that has already grown around its own dysfunction. Over years, the staff invented a hundred small human workarounds for the broken seams between systems, and those workarounds became load bearing without anyone deciding they should be. The receptionist keeps a paper list because the portal loses appointments. The biller has a private spreadsheet that is the only honest record of where claims actually sit. The nurse calls the same three patients by memory because the reminder system was never trusted. The day you try to automate any one of those tasks, you discover that the patch was holding up the building, and that what looked like a workflow was really a person quietly compensating for a design nobody chose on purpose. You can do this work. We do it every week, carefully, and it is worth doing. But you are always subtracting before you can add, untangling before you can build, and subtraction in a working clinic is painful in a way that has very little to do with software.

A clinic at day zero has no patches yet. Nothing has calcified, no one has learned to compensate, and there is no muscle memory to fight. This is the rare case where you are not retrofitting a nervous system onto a body that already has one. You are deciding, deliberately and once, what the nervous system should be. And the decisions that matter most are the ones founders are most tempted to treat as afterthoughts. The record system you choose and the phone setup you sign for in your first week are not plumbing. They are the ground every future capability will either stand on or sink into. A modern, well connected record system is the difference between an agent that can read a patient's eligibility and book against a provider's real availability, and an agent that is locked out of the very information it needs to be useful. The founder who picks those tools with the next decade in mind, rather than the next invoice, has made a choice that compounds quietly in their favor for years. The one who picks for this month has, without knowing it, set a ceiling on everything that comes after.

What changes when you build with automation present from the first day is not mainly that the work gets done faster. It is that certain work never becomes a human's job in the first place. Think about what a front desk learns to do in its opening months. It learns to answer the same handful of questions over and over, to chase patients who need to reschedule, to run eligibility checks against payers, to call the people who have appointments tomorrow and remind them to come. These are bounded, repetitive, verifiable tasks, the kind a software agent can do reliably and tirelessly, and in a clinic that opens with that capability already running, no one ever staffs for them, no one ever builds a culture around them, no one ever comes to believe that this is simply what working at a clinic means. The reminders go out from the first week. The rescheduling calls get made before anyone has been hired to dread making them. The inbound questions get answered at two in the morning by something that does not sleep, and the claim statuses get chased overnight in batches that land on a desk as a clean summary before the first coffee. The clinic never learns the busywork, so it never has to be relieved of it.

This shows up most plainly in who a new clinic hires. A practice that opens with its administrative spine automated does not hire three people to answer phones and one to chase claims. It hires one person whose entire job is judgment. The exceptions. The patient whose voice catches in a way the words do not explain. The situation that has never occurred in quite that form before and needs a human who can read it. That is a different clinic, financially and culturally, and the difference is not subtle. The established practice that wants to get there has to walk the hardest road in operations, which is telling people that the work they have done faithfully for years is now done by something else. Nobody enjoys that conversation, and the good ones avoid having to dress it up. The clinic that started clean never has to have it at all. It is far easier to never build a department than to unwind one, and the kindness of that, to the people you would otherwise have hired into work that was always going to be automated away, is real.

There is a quieter advantage too, and it pays out over the long run. A clinic that runs on structured, automated workflows from the beginning generates clean data as a simple byproduct of operating. Every interaction has a shape. Every outcome can be checked against something real, whether the appointment landed on the calendar, whether the payer returned an active status, whether the patient showed up. The founder who started this way has honest visibility into their own practice from the first month, instead of trying to reconstruct two years of history from a biller's private spreadsheet when they finally go looking for it. And because the agents learn from clean examples produced at volume, they get better faster, which means the clinic that began with automation does not just start ahead. It pulls further ahead, because good data is the thing that makes the next capability cheaper to build than the last.

I want to be careful here, because there is a version of this argument that oversells itself, and I have no interest in making it. Starting at day zero does not mean automating everything, and it does not mean a clinic can be run without people. We draw a hard line at the edge of clinical judgment, and we draw it on purpose, not because a better model is a year away. Whether a symptom warrants an earlier visit, whether a medication is right, whether a worried patient needs a human voice rather than an efficient one, all of that stays in human hands, and a good day zero design is one that protects those moments rather than crowding them out. The real advantage of building early is not that you get to remove the humans. It is that you get to decide, calmly and in advance, exactly where the humans belong, so that their time lands on the work that genuinely needs them and nowhere else. That is a decision best made with a clear head in an empty building, not in a panic two years in, surrounded by workarounds you can no longer tell apart from the actual work.

This is also why we ask to be involved before there is anything to automate. The most useful thing we do for a new practice is not flip a switch on opening day. It is sit inside the clinic while it is still taking shape, watching how the founder imagines the work, where the patients will come from, how a visit will actually move from a phone call to a filled chair, and helping compose the administrative design around all of it rather than bolting capability onto a structure that has already set. There is a habit I have kept from a different life, spending real time inside a place before presuming to build for it, and it matters most precisely here, in the weeks when everything is still soft enough to shape. You cannot ethnograph a workflow that has already hardened into folklore. You can only excavate it. But a workflow that does not exist yet can be designed well the first time, and designing it well the first time is worth more than any rescue we could mount later.

The founder who calls before opening is not more cautious than the rest. If anything they are more ambitious, because they have understood something the firefighting clinics learn too late, which is that the administrative shape of a practice is not a cost to be managed after the fact. It is part of the medicine. A call that never gets returned, an eligibility check that never gets run, a reminder that never goes out, these are not back office failures. They are how patients quietly fall through, and a clinic decides, in its very first weeks, how often that will be allowed to happen. The empty clinic is the one chance to decide it deliberately. It does not stay empty for long, and the founders who use that window well end up with the practice they actually wanted in the first place. The rooms fill with patients instead of other people's habits, and the people who work there spend their days on the part of the job that only a person can do. That was always the point of opening the doors. It is a strange thing, that the way to get there is to think hardest about the work before any of it has begun.