236: Reid Health's CDIO Muhammad Siddiqui: Being 2nd to Go-Live with Epic’s Emmie AI, Epic Gold Stars 2 to 10 in Six Months, and How AI Can Scale Dysfunction if You're Not Careful

Subscribe on: RSS | SPOTIFY | APPLE PODCAST | GOOGLE | BREAKER | ANCHOR

On this episode of The Digital Patient, Joshua Liu, MD, Co-founder & CEO of SeamlessMD, and colleague, Alan Sardana, chat with Muhammad Siddiqui, Chief Digital and Information Officer at Reid Health, about "Being 2nd to Go-Live with Epic’s Emmie AI, Epic Gold Stars 2 to 10 in Six Months, How AI Can Scale Dysfunction if You're Not Careful, and more..." Click the play button to listen or read the show notes below.
‍

Audio:

‍

Guest(s):

• Muhammad Siddiqui, Chief Digital and Information Officer at Reid Health
• Joshua Liu, MD (@joshuapliu), Co-founder & CEO at SeamlessMD

‍

‍Episode 236 - Show Notes:

[00:00:07] Episode preview

[00:04:50] The context that shapes every technology decision at Reid Health — a not-for-profit system in Richmond, Indiana serving both Indiana and western Ohio, with roughly 3,500 people, one hospital, and several outpatient clinics across a mostly rural and semi-rural region. Muhammad is blunt that they are not a Cleveland Clinic or a Mayo Clinic and don't have the bench depth or the finances, "but the mission just as real, right? Maybe more urgent," because there is no backup option nearby — for a patient having a heart attack or a stroke, the nearest other urgent care center is 60 miles away.

[00:06:10] Why Reid Health doesn't wait for ideas to become perfect — workforce pressure, margin pressure, and clinician burnout show up every single day, and sometimes faster than they do for larger systems. The approach from day one was to avoid chasing technology because it's new, flashy, or because everyone else is doing it, and instead to look for pain points where technology removes real friction: "Don't chase AI, chase the problem to solve with the AI." With Abridge, the question was never whether they could be first, second, or third — it was whether they could give time back to providers and improve the experience inside the room.

[00:07:14] How he frames time as the currency the health system is really spending — a patient waits 15 minutes on the phone with the contact center, waits again in the car, waits for parking, waits inside the hospital, and then finally sits with a provider who isn't focused on them. "The last thing we wanna do, a patient is sitting with a provider in the room, and provider is no focus and not really focusing on the patient. For me, that's a problem as well, and that was a problem we tried to solve there as well."

[00:07:54] What "staying close to the work" means in practice, and why a rural system can be more agile than people assume — Reid Health doesn't treat AI as a lab experiment. It goes inside the real clinical workflow, with the providers who will actually use it involved from day one, and with real patients through a strong PFAC committee and patient experience team, which lets the team learn very fast what problem the tool is actually solving. The second advantage is speed of decision: "My CEO sit next door to me, I can just knock the door and walk in there. I don't have to go through 15 different committees to get their approvals." That made ambient AI a no-brainer two and a half years ago, and Reid became account number two for Abridge after Kansas City. By day five or six, doctors were knocking on his door asking when they could get access. The lens behind all of it: AI isn't about looking advanced, it's about solving the problem that matters at 3:00 AM, when the patient is in bed, the nurse is tired, the provider is behind on notes, and the system still has to work.

[00:10:30] The reason inpatient nursing ambient AI is a fundamentally harder problem than physician documentation — a physician visit has a clear structure: conversation, assessment, plan, note. Inpatient nursing is continuous and fluid, with nurses moving between patients, families, medications, handoffs, alerts, and interruptions, sometimes at five- or six-to-one ratios. "So challenge is not just the technology. The challenge is understanding the nursing workflow well enough" — and delivering support without adding a single task to an already overburdened group, because the moment you do, adoption collapses and trust is gone.

[00:12:10] How he opened the nursing kickoff, and what came back from the floor — his message to nursing leadership was explicit: this is not surveillance, this is not extra work, there is no additional screen or keystroke; in fact the technology removes keystrokes. The team spent real time with nursing leadership and at the bedside asking practical questions, and even asked patients how it would help them, running a detailed requirements elicitation alongside Abridge and working through the Epic integration. Rounding on the fourth floor the day before, a nurse thanked him directly: she used to spend an hour after her shift on documentation, and can now go home on time and be involved in her son's scouts this year.

[00:16:35] Why other health systems are now coming to Richmond to learn — not just startups, but a Mayo Clinic team from the Arizona campus visiting to learn about Reid's smart room technology, and Parkview Health and Kettering Health bringing their leadership to see how Reid implemented Abridge and its smart rooms. "Technology become an equalizer... we really don't want a zip code to dictate the quality of care," he says, framing it as the IT team's mission to give a vulnerable community the same technology advancement available at any other health system.

[00:18:11] Why patient-facing AI inside MyChart earns more trust than a consumer LLM — Reid was the second health system live with Emmie, Epic's patient AI in MyChart, giving patients plain-language answers about test results and visits. Muhammad is clear he isn't anti-GPT and uses AI himself "to learn, not to become a lazy person," but healthcare is different: when a patient has a test result or medication question, the answer has to be connected to their own clinical context, not a general answer from Dr. Google. Patients already trust MyChart as their own information, and the ambient conversation from the exam room lands in that chart — context a general-purpose chatbot will never have. What patients want isn't medical jargon: "What does this mean for me? Should I be worried? What did my doctor say? What should I do next?"

[00:21:33] How a single "abnormal" lab result illustrates the risk of context-free AI — paste that result into a general chatbot and it will likely call it abnormal without knowing the patient's medications or the trend, when the value may in fact be improving. "In healthcare the safest AI is not the smartest-sounding AI... It is the AI grounded in the right workflow and the right clinical guardrail."

[00:22:47] Why most health systems aren't behind on AI, they're behind on the basics — AI won't fix a broken workflow, it will make a broken workflow run faster. Inconsistent scheduling templates won't be magically fixed by AI scheduling; weak data governance won't magically produce clean insight; and if clinicians don't trust the tool, AI won't manufacture that trust. When he takes this to the board and executive team, he connects the basics to risk and value and reframes it: this isn't slowing down or adding bureaucracy, it's moving in the right order so the organization doesn't waste money, damage trust, or create operational confusion. "Don't just put that AI on top of your messy workflow and you will be fail from day one."

[00:25:08] How Reid Health sequences an early-adopter rollout with no playbook to copy — nursing started on a single floor in phase one and expanded to the whole organization within 60 days because the results were phenomenal. The provider rollout followed the same pattern, starting with one medical informatics director who was also his own PCP; she became the brand ambassador who helped scale it to many providers overnight. Abridge is now used by the majority of Reid's providers, including in the ED and urgent care, is being tested in the OR, and has been extended to MAs and nursing in ambulatory care centers.

[00:26:29] The painful early-career lesson that changed his definition of success — "One of the biggest mistake we make in a technology is thinking that go live means success." He has seen systems go live that worked technically but made the day harder for the people using them: more clicks, more alerts, more workarounds, more confusion, while IT celebrated delivering on time and on budget. His example is Epic Refill two and a half to three years ago, when Reid was three versions and 700 features behind: they implemented all 700 features, but adoption stayed low because they turned on functionality without fixing the underlying workflows. Now the questions are different — did we make the work better, did we reduce friction, did we give time back, did we improve safety and the experience for patients and caregivers?

[00:29:30] What his adoption numbers look like across 21-plus live AI tools, and his advice to the vendors building them — Reid is live with more than 21 AI tools across Epic and Workday, and he is candid that adoption sits somewhere in the middle rather than high, and that maturity will take time. Some of the features are amazing, but his recommendation to startups is to go where the action is: "Do not start building a product that just showcase the value on a paper rather than in ground reality."

[00:30:35] Why the demo is not where healthcare happens — the underlying workflow and persona haven't changed, and the goal isn't to digitize a process so much as to have it run on the back end without anyone noticing. He sees polished, smart-looking demos every week, but "healthcare happen in messy workflows" — staff shortages, patient complexity, payer rules, and processes nobody has touched in years. During a patient access project two years ago, Reid found CT and mammography scheduling templates that hadn't been touched in eight years. That's why Reid built a system performance improvement department reporting into the CIO office, running Lean and Kaizen work to understand the current-state and to-be process first, then finding the tool that solves it.

[00:33:51] How Reid Health went from Epic Gold Star level 2 to level 10 in under six months — the Epic Refuel playbook he now shares with other organizations came down to four things. It was treated as an organizational effort, not an IT project, with a physician chairing it rather than IT. Provider ownership was real: IT didn't sit in a room deciding what to turn on. The guiding principle was Epic-first but not blindly Epic — staying close to foundation to avoid heavy customization. And transparency, via an Epic Refuel ROI dashboard showing monthly value: self-scheduling went from under 1% to roughly 4–5% within 90 days. Reid has held Gold Star 10 for two years running.

[00:36:45] What separates genuine provider ownership from performative stakeholder engagement — technology leaders instinctively organize the work, define the plan, manage risk, and drive the timeline, but "if clinicians are going to be truly taking the lead on that project, IT need to stop acting as the owner of the agenda." IT's role shifts to translator, advisor, and builder: explaining what's possible in Epic, handling build, security, data, training, and change control, while clinical value gets defined by the people doing the actual work — the ones in the exam room or walking into a patient's room to talk with a family member.

[00:38:51] The four questions Muhammad puts to every project before it moves — does it actually solve a real problem affecting patients, access, margin, safety, or risk, not a theory or a white paper? Do we have the foundation — workflow, data, governance, security alignment — and if not, what has to come first? Can we measure the value, including soft ROI like time saved, risk reduced, access improved, or burden lifted, defined before you start? And what happens if we do nothing, and what's the opportunity cost of that? He notes the framework doubles as a filter for internal requests, and for a rural system it's essential: "We cannot afford technology that looks good in a board deck but does not change the work."

[00:42:20] Where healthcare leaders should deliberately slow down — not on speed of delivery, but around trust and around broken processes. Success should be measured on adoption and ROI rather than go-live, and clinical trust, patient trust, data trust, and operational trust all have to be built. "You can roll out technology overnight. That's not a problem. But if people don't trust, they will work around it" — and once that happens, it's hard to recover. The same goes for automating dysfunction: "If the workflow is broken, AI may not fix it. It may scale the dysfunction." His starting questions for any project: what problem are we solving, who owns that workflow, what does success look like, what could go wrong, how will we know if this is helping, and who has to change the way they work today?

‍

‍Fast 5 Lightning Round:

1. What is your favorite book or book you’ve gifted the most?
   The Five Dysfunctions of a Team by Patrick Lencioni
2. If you could instantly master any skill, what would it be?
   "Storytelling"
3. Would you rather have Super strength, super speed, or the ability to read people’s minds?
   "Mind reading."
4. What is something in healthcare you believe others might find insane?
   "I think the small and midsize rural health systems can out-innovate much larger organizations."
5. What is the last movie or TV show you saw, and what did you think of it?
   "The Pitt."

The Digital Patient has been recognized as Feedspot's #1 Patient Engagement Podcast of 2025. Thank you to our listeners for making this happen!