Written by Paul Brient, Chief Product and Operations Officer, athenahealth

I wrote my first healthcare software when I was still in high school. My father had left academic medicine to open a solo surgical practice, and billing meant someone had to sit in front of a typewriter and create claims one by one. Painstaking. So I taught myself to automate the process using a computer. By the time I left for college, my software was running in more than 150 physician practices.

Nobody had to force those practices to buy it.

That was a win I didn’t fully appreciate at the time. Practice management software was adopted willingly because it was useful. The software helped their staff run the business of medicine, and so they paid for it out of their own pockets.

The electronic health record (EHR) is another story. By 2008, practice management systems were already widely used by physician practices, yet just 4% of U.S. physicians reported having a fully functional EHR. It took the HITECH Act and Meaningful Use , with billions in government incentives, to change that. Today, roughly 95% of office-based physicians use an EHR.

This shift created something undeniably valuable, a digital clinical record at scale. However, despite these mandates, the tools weren’t ready for practice prime time. We digitized before we redesigned, taking our burdensome paper-based processes electronic without streamlining the workflow first. This also placed an inordinate burden on the physician, who became responsible for entering and structuring the data to make the tools work. In 2016, researchers observing ambulatory physicians found they spent nearly twice as much of their day on the EHR and desk work as they did face-to-face with patients.

So, physicians got creative. They hired medical scribes to follow them from room to room, taking notes and operating the computer. They bought third-party reporting and insights dashboards to get answers the system couldn’t give them. In many orthopedics practices, medical assistants fill out a single-page paper form with the relevant patient information and hand it to the doctor. The doctor takes it into the room and is patient-focused and tech-free. In academic medicine, attendings on rounds rarely touch a computer, because there are residents to do that.

There is a level of absurdity here. Organizations pay millions of dollars for these systems, mandate days of training, and staff entire technology teams to stand them up and maintain them. And then some of the users those systems were built for hire other people to operate them.

From the outside, it looked like physicians were resisting technology. In retrospect, we had it backwards — the scribe wasn’t just a workaround; it was a description of what the physicians wanted from their software.

The scribe provided a layer of interpretation between what was happening in the exam room and what needed to be entered into the computer. The fundamental problem with EMRs at the time is that they were “dumb.” They took characters that were entered by physicians, stored them, and faithfully brought them back when asked. They didn’t “understand” the data, and without a lot of programming and customization, couldn’t really assemble it in a tailored way for providers. At many levels, EMRs were data entry forms with a nice filing system and a search hierarchy.

When you ask physicians what they’d want out of their ideal EHR, they invariably describe the same thing: one screen that has everything that is relevant for the current situation (and nothing more). They don’t want to spend precious time clicking around, hunting for the information they need and having to assemble it themselves.

The challenge, of course, is that for any given patient, doctor, and situation, what matters is different. A cardiologist and an endocrinologist seeing the same patient on the same afternoon do not need the same information in front of them. An orthopedist working a patient up for a total joint replacement needs very different information than when they see the same patient six weeks after surgery.

We would have loved to build this years ago. We simply couldn’t.

That has changed. AI is the first technology that puts the physicians’ request within reach. It allows us to build a system that can “understand” the patient, provider, and context of the visit and provide a summary — much like what a physician familiar with the patient might do. It can surface the relevant clinical information that matters for a given visit, find the relevant line from last month’s follow-up note from the specialist, listen to the patient-provider interaction and, based on that, draft a note, prepare orders, and assign appropriate billing codes. Done right, the EHR becomes the best medical assistant, or the best resident you’ve ever worked with.

By McKinsey’s count , approximately 2,400 AI-enabled healthcare companies launched between 2015 and 2024, most of them building something that sits on top of a legacy EHR that can’t reason. It is an improvement on what came before, but it creates its own set of complexities and challenges — robbing physicians of finally having technology that truly works for them. I equate the current moment with the dot-com/eHealth era of 1999 — tons of startup companies doing things that were better than pre-dot com but fell well short of the ultimate expression of what was possible. The thousands of dedicated online retailers gave way to the Amazons and eBays — companies that truly figured out how to reimagine retail using the technology rather than simply take orders on the internet.

At athenahealth, we’re reimagining the EHR and the practice management system from the ground up without requiring our customers to do anything other than just enjoy the new technology. Our goal is an EHR that requires minimal touching of the computer and is so easy to use it is “user obvious.” We are getting out of the business of training physicians on how to use software. We want something that feels more like the Uber app. You don’t think much about it — nobody shows up to train you, and it works the way you expect. That is what the EHR has to become, and it is a long way from where any of them are today.

Yet it’s easier to imagine than ever before. Here’s what finally tells me we’re on the right track: nobody is mandating adoption of AI. Physicians and nurses are adopting AI tools on their own. In a single year, repeated daily AI use among physicians tripled, from 10% in 2025 to 38% in 2026, while use among nurses doubled, from 16% to 32%.

The other thing that has changed is how this software gets built. The first generation of EHRs was designed largely against regulatory requirements, and clinicians had far less influence over how those systems were built than they should have. That is no longer how we work. Validating clinical AI is genuinely hard. The only method I trust is a practicing physician assessing the model’s work on a patient they actually know and remember. That’s why we created the EHR AI CoLab, a co-innovation forum within athenaInstitute , athenahealth’s research and intelligence hub. Made up of nearly 20 clinicians and EHR superusers, the CoLab brings end user feedback into the evaluation, testing, and refinement of new AI capabilities.

Getting to where we want to go requires a change in mindset. In the past, when we had a confusing screen, we’d add a tooltip, create a Pendo, or, if it was really confusing, create a training video or course. In-app guidance is not the answer. Neither is bolting on an assistant. That’s why what I’m after is the EHR we should have built in the first place: fully reimagined around what AI now makes possible. Somewhere along the way we decided it was normal to send physicians to several days of training before they could use a system meant to help them take care of patients. It isn’t normal, and it’s not going to survive this. The EHR that wins the next decade is the one nobody has to be taught to use.