Unless you’ve been living under a social media rock, you know that Epic held its User Group Meeting (aka UGM) this past week. A decade ago, it was a place to see cool development but the buzz was a bit muted. I recall puttering around in the evenings with my team trying to figure out where to eat dinner.
No longer. Now Verona becomes a kind of healthcare Disneyland every August, complete with the big-stage spectacle, polished product demos, and a Jobsian “one more thing” moment that feels straight out of Cupertino. If you want a list of new features, there are plenty of people posting screenshots and bulleted recaps. This isn’t that. If you want a recap, I would first ask where you’ve been over the past week and then feed you these links:
https://www.healthcareittoday.com/2025/08/21/a-deep-dive-into-the-announcements-at-epic-ugm-2025/
https://www.newsweek.com/epic-ugm-2025-live-updates-verona-wisconsin-2115581 (I thought this was especially good for non-Epic folks but honestly, I didn’t even think that Newsweek was still a thing.)
What I find more interesting are the quiet undercurrents. What is some of the connective tissue between what Epic shows us on stage and what it might actually mean for how medicine is practiced.
The Promise of Cosmos (and Why It Matters)
If you’ve followed me in the past, you know that I’ve been a Cosmos fanboy before it was cool. Back in 2022, I posted about mortality differences with COVID’s Omicron variant and felt like I was looking at the future but recognized that there was lots of potential, not much usability. The interface was stiff and the query options limited.
Fast forward three years and it’s becoming a genuine suite of tools that is clearly the core of what Epic will be for the foreseeable future. This matters because medicine’s dirty secret is that evidence translation takes 20 years. In other words, it typically takes twenty years from discovery to standard practice. By the time new knowledge trickles into actual care, it’s not new anymore. And this doesn’t even add in the time it takes to come up with the discovery or evidence in the first place.
Epic’s constellation of AI-infused tools hints at breaking that cycle. If Cosmos + AI can shrink evidence translation from decades to months—or even weeks—that’s not an incremental step. That’s a new axis of time.
Registries Reimagined
One of the most gratifying things I’ve seen is Epic’s shift toward Community Registries, an idea I talked about on stage at UGM in 2022. Specialty registries matter. They are used for accreditation, benchmarking, and quality improvement. Unfortunately, the burden of connecting to them has always fallen on overstretched IT teams, dedicated abstractors and already burned-out clinicians.
The Community Registries approach instead leverages some obvious entry points in Epic. Instead of custom interfaces and manual data entry, Epic is baking in downloadable forms, automated submissions, and lifecycle support. Less duct tape. Less swivel-chair integration. More value, faster.
This is the kind of friction-reducing infrastructure that rarely gets applause but quietly changes everything.
The AI Shift: From Words to Events
Much has been made of Emmie, Art, and Penny—the mascots of Epic’s AI push. What excites me isn’t the branding; it’s the move from text-based predictions to event-based reasoning.
Instead of asking, “What’s the next word?” Epic’s large medical language model asks, “What’s the next event?”
A poorly controlled heart failure patient → admitted for medication management → multiple medication changes → discharge… the next event isn’t prose or an email. It’s readmission.
Now imagine layering “what if” data. What if we add pharmacists for reconciliation? What if we adjust follow-up intervals? Suddenly, you’re doing digital twin simulations with real-world plausibility. That’s not hype—that’s system redesign at machine speed.
The Invisible Middle Layer
Epic was clear about their three priorities for GenAI: patients, physicians, and financial health. Fair enough. But I couldn’t resist asking: where’s the love for the informaticists and analysts?
Because here’s the real unlock: imagine if analysts could “vibe configure” Epic. Feed it a new guideline PDF and get back order sets, print groups, decision support, and dashboards—all aligned and live within weeks, not years.
That’s the missing middle layer. Not the patient-facing chatbot, not the executive dashboard—but the connective tissue that translates knowledge into workflow. If Epic can empower that, it changes the economics of change itself. Fortunately, I got news that this is indeed on their development timeline. It didn’t get highlighted on the big screen but it is reassuring to someone like me who will hopefully make heavy use of it in the future.
Hype, Pace, and Cold Reality
At every UGM, someone inevitably asks: “Is there anything Epic isn’t trying to take over?” The answer: no. And even when Epic does move in, the pace is slower than you think. Remember when Secure Messaging was supposed to end PerfectServe and TigerText in two years? They’re still around and many organizations are still clinging to them like pagers and fax machines.
Healthcare change management is not a sprint—it’s a slow, bureaucratic slog. Epic can ship features; organizations still have to absorb them. If you doubt me, I have three words for you: Value. Based. Care.
For vendors in adjacent spaces, the window is narrower but not shut. For instance, if you’re in ambient documentation or workflow AI, you don’t have five years to find your niche. You have more like eighteen months. Your edge isn’t competing with Epic’s breadth—it’s outpacing their depth in focused use cases that make your tool invaluable to the organizations that use it.
What Wasn’t Said
For all the talk of AI, registries, and digital twins, I was struck by what didn’t get attention: data governance and content quality.
These are the unglamorous, grinding projects that determine whether any of the flashy tools will actually deliver value. Governance is like plumbing—nobody celebrates it, but without it, everything leaks.
The operational and political hurdle is that the cost of governance accrues to IT and informatics, while the benefits show up in operations and outcomes. That misalignment makes it chronically under-resourced. Yet without it, every promise on the big stage risks collapsing under its own weight.
Final Reflection
Much has been made of AI replacing healthcare jobs. A hospital CFO attending UGM might start secretly drooling about all the FTEs they can start RIFing. I would counsel that CFO on both getting a soul transplant but also on the concept of Jevon’s paradox, which tells us that more efficiency often creates more demand, not less. I think this will occur in spades and will benefit organizations that think of these tools beyond a zero-sum game. For instance, automated coding won’t eliminate coders; it surfaces an opportunity to use their skills in roles like data abstraction, registry curation, and quality assurance, all of which I think will be subject to Jevon’s paradox.
The real risk isn’t job loss. It’s that we ignore the opportunity to leverage these new technologies by dedicating resources to the infrastructural work that allows any of this to matter. Evidence translation. Governance. The analyst layer. These are the quiet levers that make transformation real.
Epic dazzled us again at UGM. But if you want to know where the future is really being written, it’s not on the big screen in Deep Space. It’s in the grinding, invisible work underneath.
> The Invisible Middle Layer
Not exactly what you're asking for but they did announce:
- MyChart Builder Agent
- Slicer Dicer Agent/Copilot
- Campaign Agent
Which all hit some of those analyst or non core roles
Excellent as always. Insightful middle layer comments and analysis. Great to see you in Madison !