235: Beacon's CIDEO Dr. Stacey Johnston: Pushing Oracle to Its AI Limits, The Agent That Ordered 7,000 Colon Screenings and Caught Cancer Early, and What If You Spent More on Tech Than on Recruiting?

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 Stacey Johnston, MD, MHA, CHCIO, Chief Information and Digital Execution Officer at Beacon Health System, about "Pushing Oracle to Its AI Limits, The Agent That Ordered 7,000 Colon Screenings and Caught Cancer Early, What If You Spent More on Tech Than on Recruiting, and more..." Click the play button to listen or read the show notes below.
‍

Audio:

‍

Guest(s):

• Stacey Johnston, MD, MHA, CHCIO, Chief Information and Digital Execution Officer at Beacon Health System
• Joshua Liu, MD (@joshuapliu), Co-founder & CEO at SeamlessMD

‍

‍Episode 235 - Show Notes:

[00:00:07] Episode preview

[00:06:27] Why she took the CIO path — for Stacey it has always been "about being in the right place at the right time, and just saying yes to new opportunities." She became a CMIO straight out of residency as the only person who had used any form of electronic order entry, while simultaneously serving as a medical director, which gave her an operational background alongside the technology framework. She then doubled down on informatics training and on challenging why things are done a certain way: what is the best workflow, not putting in clicks just because you can, and not using alerts to teach provider behavior.

[00:07:37] How leading a big-bang Epic implementation redirected her career — recruited to Baptist Health in Jacksonville as CMIO, she was then asked to lead the move to Epic, her first after leading Cerner and Meditech conversions. Epic's integrated revenue cycle meant working across operational and clinical leaders on a two-year, organization-wide transformation rather than the slow, one-office-at-a-time roll of prior EHR conversions. Owning cloud hosting, infrastructure, and Hyperspace, she realized she genuinely enjoyed supporting the technology — and built out the digital-first strategy and digital front door alongside marketing and consumer teams.

[00:09:51] The accolade she valued most out of that implementation — the team collected the Good Install Award, Epic stars just 11 months after go-live, a Summa Cum Laude Honor Roll placement, and CHIME's Most Wired award. But her favorite was the employee engagement survey, where Epic rated number one across all 67 questions. That signalled the work was done: "I'm like Mary Poppins. I came and you needed me, and now you no longer need me."

[00:10:58] What made Beacon the right next move — the organization was further behind in technology than it wanted to be, but leadership was ready to become a digital-first organization and needed someone to create the vision and get them there. A core part of the mandate was turning EHR and technology vendors into genuine partners "instead of technology being reactionary."

[00:12:04] How she stays close to clinical informatics as a CIDEO — she is a believer in skip meetings, meeting directly with the managers and team leads over informatics and clinical applications. There had been no IS governance in place, so she brought over what worked at her prior organization: an IT steering committee, eight advisory councils, and work groups reporting into them. Beacon is also building a true physician informatics program beyond a CMIO and a single medical director, having just recruited a medical director of informatics for the surgical space.

[00:13:18] Why IT is being realigned to service lines — instead of technology teams thinking "this is my technology, and my technology is good," Stacey wants a named cardiology person who attends cardiology operations and monthly departmental meetings and handles everything from the EHR to the cardiology PACS to device monitoring, so clinicians don't chase three different people to fix one problem. It's a two-to-three-year journey requiring the right hires, and clinical informatics is already service-line assigned. The same logic drives hiring: Beacon brought in a nurse practitioner who worked for one of its neurosurgeons as manager of IS over surgical services. "We can train you on the technology side, but it's much harder to train someone on the workflows or the clinical workflows."

[00:18:19] What AI means for the vendor landscape, and the federated model underneath it — she still expects occasional niche vendors but fewer of them, with health systems doubling down on their core partner. As an Oracle beta partner, Beacon pushes Oracle to develop AI that is meaningful and usable, while Oracle pushes Beacon to stay on the latest code and do its upgrades. Because Oracle won't meet every need, a medium-sized system ends up hybrid: building large language models doesn't make sense outside a major academic medical center, but Beacon runs an AI citizenship program where associates build simple low- and no-code automations with Copilot, UiPath, and similar RPA tools. Under that federated model, simple agents get built at the individual contributor level, more complex work is centralized with an internal AI team, and very complex workflows or true digital employees go to carefully chosen partners vetted on compliance, where data is stored, how they monitor for drift or bias, and their ethical considerations. Beacon's AI governance program was recognized by AVIA, and the system is now implementing an AI monitoring platform after demoing several.

[00:21:17] Why the technology-versus-talent budget split should flip — healthcare IT budgets run about 4 to 5% of net revenue, while talent and acquisition spend runs 5 to 8%. Stacey's question: what if you spent 4 to 5% on talent and recruiting and 8% on technology? "We can't even fill the seats we have. There just aren't enough people to work in healthcare anymore." Rather than hiring five people, hire two and elevate their skill sets with AI or have them monitor a digital employee — but don't implement AI for AI's sake, and go back a year or two later to confirm the ROI actually materialized.

[00:24:04] How "fail fast" works in practice — her AI team's mandate is to turn on an agent in three weeks, monitor it for six months, and turn it off if it isn't meeting the need. The in-house RPA team does the same with bots that take more time and effort than expected. The proof point: when Beacon acquired four hospitals in southwest Michigan and went live on February 1st with a backload of 100,000-plus appointments, the complex acute appointments were staffed with people, but the repeatable ambulatory backload would have required 40-plus extra hires. A vendor built an agent in three weeks; about a week after go-live it had self-corrected and taught itself to schedule, ultimately booking 70,000 appointments in three weeks. Then they shut it off. "You have to be willing to throw away some work."

[00:26:08] How an agent closed a colon cancer screening gap and caught a cancer — running behind on colon cancer screening ahead of the December fiscal year-end, the quality team planned to mass-sign paper Cologuard orders, which would have returned results as paper PDFs. Stacey pushed for an agent to place true electronic orders so results came back as discrete data. Built in about three weeks, it ordered 7,000 screenings; 200 returned positive, nearly that many colonoscopies followed, and one early colon cancer was detected. The agent was then modulated from a year-end catch-up into ongoing monitoring: when a patient turns 50, it notifies them, places the order, ships the kit, and returns discrete results.

[00:29:00] Where voice agents come next — today's agents are text-based, but Beacon is moving toward voice in HIM and case management. The filter is which repeatable calls are going unstaffed today: patients who get three text attempts and then a manual phone call, case management follow-ups on whether an appointment happened or got scheduled, and HCAHPS surveys that pull charge nurses away from patient care. "These are these simple repeatable processes is what you wanna start with first. Do the low-hanging fruit."

[00:31:30] Why she insists on foundation before ROI — in her first week at Beacon in October 2024, she presented to roughly 200 board members and physicians, and the first two slides of her vision deck said she needed a good foundation first: data cleanup, standardization, foundational partners, and application rationalization, since multiple vendors doing the same thing drives maintenance cost and disparate systems and data. The hardest and most important piece is workflow standardization, where being a physician leader helps because she knows how hard it is to move clinicians' cheese. Her guiding principles carry that argument: put the patient and family first, then enhance the caregiver experience. "Is it bringing joy back to the workplace? If you're doing those things, through standardization, then the pushback that you're gonna get about standardization is gonna be minimized."

[00:33:56] How ambient AI became the early win that built trust — coming from an Epic-first organization into an Oracle-first one let her say she'd seen the other side and still believe in this partner, while holding Oracle accountable for the product and letting Oracle hold Beacon accountable for participating in build decisions. Beta partnership is real work — weekly R&D meetings, Oracle on site with the physician informatics and build teams — but it delivered the clinical AI agent, Oracle's fully embedded ambient listening tool. After early bumps outside primary care and in the inpatient setting, roughly 70% of ambulatory physicians now use it and 70% of those physicians' notes run through it. Next month brings Oracle's next-generation EHR in ambulatory pilot sites with an AI semantic layer for summarization, dictation, search, and chat that will queue up orders alongside the note. On the nursing side, wins came from care and feeding of core systems: event care sets and learning to document by exception.

[00:36:49] Why a physician CDIO evaluates vendors differently — with deep clinical experience she can get into the weeds on a demo that would otherwise look great, and she'd just had one the week before where her questions kept landing on "we're working on that." Her verdict: come back in a couple of months, not quite ready for primetime. Even so she stays deliberately vendor-friendly, meeting startups and taking pilot and beta roles, because you never know where a good partner will come from. Being a beta partner everywhere is exhausting for the team, but the right ones are energizing — a reminder that IT teams carry cognitive burden too, and a job of only upgrades and tickets holds little joy.

[00:38:41] Where to be an early adopter and where to sit back — she challenges application leaders to be at the leading edge of their core EHR: study the roadmap, ask what isn't turned on that should be, ask to see the AI list, and push to embed it into workflows. That requires care and feeding — a regular upgrade schedule and standardization toward a model or foundational build so upgrades and new functionality land more easily. Beyond the EHR, target where you're operationally struggling: if your length of stay is already strong, don't invest in AI aimed at length of stay. Beacon's targets are the roles it can't fill, which is why it is doubling down on digital associates and voice call AI.

[00:41:21] How 150 AI use cases got sourced from the front line — at a recent leadership connection day, Stacey hosted a tabletop exercise with nearly 500 participants across roughly 50 tables, each with 15 minutes to produce one AI use case, with Beacon swag for the winner and a commitment to build the winning idea. It expanded into open individual submissions and grew from 50 ideas to 150. Sifting them meant separating "auto magical thinking" — one thing that does a million things — from what's realistic now, then assessing the build effort, workflow change, change management, education, and whether the systems exist in-house or a partner is needed.

[00:42:47] What Beacon's AI policy permits to run autonomously — fully autonomous AI has a place, but on the back end, more in revenue cycle, and only with QA checking and someone monitoring the system, the way RPA bots are already monitored by a team watching for anything amiss. Leaders need to be trained to monitor their agents: "you don't just set it and forget it." On the clinical side the policy holds that AI augments the clinician's workflow rather than replacing it — radiology AI modules queue up the studies to look at first without excusing the rest, and ambient listening queues the document but the clinician still reads and signs it. She wouldn't be surprised if autonomous clinical care comes, but says the system needs to be very deliberate about where.

[00:45:04] The advice she'd give her younger CMIO self — find the balance between risk-averse and risk-taker. "If you're always a risk-taker, your failures will lead to ultimately, you know, potentially a patient safety event. If you're always so risk-averse, you're not gonna move forward." She was more risk-averse early on and has since moved toward agile methodology and failing fast. The other lesson: knowing how to bring people along and understand their concerns, because "you could have the best technology in the world, but if your people aren't gonna use it because they're not there emotionally, mentally, from a training standpoint, they're not gonna use it" — EQ has to back up technical acumen. Her closing advice extends that to herself: keep trying new things and keep taking classes and attending conferences, currently an AWS fellowship and a certified physician executive program, because "the more I can better myself, the more I can bring that betterment to the organization" — and because meeting peers doing great work beats replicating the wheel yourself.

‍

‍Fast 5 Lightning Round:

1. What is your favorite book or book you’ve gifted the most?
   "My husband (T.D. Johnston) is an author, so if I didn't say his new book The Furtherer, he may have to say something... But also Pride and Prejudice and Where by Jane Austen, and Where the Wild Things Are by Maurice Sendak."
2. If you could instantly master any skill, what would it be?
   "Navigating multiple things. I mean, I'm pretty good at multitasking, but maybe being more purposeful with that."
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?
   "The number of pages of pieces of paper we're still printing and scanning on a daily basis is insane."
5. What is the last movie or TV show you saw, and what did you think of it?
   Young Sheldon with my son.

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