228: UToledo's former Enterprise CMIO Dr. Ryan Sadeghian: Don't Outsource Your AI Thinking, The Reason Clinical AI Tools Die After the Pilot, and Was Documentation Ever the Right Measure of Care?

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

On this episode of The Digital Patient, Dr. Joshua Liu, Co-founder & CEO of SeamlessMD, and colleague, Alan Sardana, chat with Ryan Sadeghian, MD, MBA, MSc, former Enterprise Chief Medical Information Officer at the University of Toledo, about why you shouldn't "Outsource Your AI Thinking, The Reason Clinical AI Tools Die After the Pilot, Why Nobody Is Auditing Whether Documentation Was Ever the Right Measure of Care, and more..." Click the play button to listen or read the show notes below.
‍

Audio:

‍

Guest(s):

• Ryan Sadeghian, MD, MBA, MSc, former Enterprise Chief Medical Information Officer at UToledo Health
• Joshua Liu, MD (@joshuapliu), Co-founder & CEO at SeamlessMD

‍

‍Episode 228 - Show Notes:

[00:00:07] Episode preview

[00:05:39] Why Dr. Sadeghian became a translator between medicine and technology — after recognizing that the hardest problems in healthcare were rarely clinical alone, Dr. Sadeghian realized they were simultaneously operational, technical, behavioral, and clinical. That insight pushed him into informatics and eventually AI, where his role became "less about choosing between medicine and technology and more about translating between the two."

[00:06:15] How the skills of sales became central to clinical informatics — convincing clinicians and staff that a new technology won't add layers of discomfort is, in Dr. Sadeghian's view, the biggest sale a CMIO makes. His background in IT sales gave him a framework for change management that he applies to every implementation.

[00:06:57] What Dr. Sadeghian's vendor governance model looks like at the University of Toledo — all clinical IT that touches the patient runs through the clinical informatics department, giving Dr. Sadeghian's team the leverage to negotiate contracts, renegotiate existing agreements, and replace vendors when needed. He also sees near-term potential for using AI tools like Claude to review contracts with a trained set of institutional rules and legal parameters.

[00:08:57] How a peer-driven "trusted advisor" model moved clinician engagement from 18% to 97% — rather than driving adoption top-down, Dr. Sadeghian built a governance layer of 67 trusted advisors across four pillars (clinical, operational, non-operational, and nursing). Each advisor acts as a peer representative between their division's leadership and IT — implementing change themselves, with Dr. Sadeghian's team in a supporting role rather than forcing a project on them.

[00:09:57] How the Nabla rollout grew to 150+ physicians through peer word-of-mouth alone — Dr. Sadeghian piloted Nabla with 40 physicians across specialties and let them spread it organically. "I do zero marketing because all we did was a pilot and get the physician to spread the goodness." Physicians now approach his team asking to join after hearing about it in clinic.

[00:11:27] Why measuring AI adoption looks different for every specialty — psychiatry and radiation oncology value finishing their day without documentation anxiety; high-volume cardiologists value the 60 minutes saved across 30 patients; GI and procedural specialties care most about documentation quality that prevents payer denials. Dr. Sadeghian tailors the value proposition to each group's specific pain points.

[00:16:26] What made 45+ specialty-specific clinical GPTs indispensable — the tools that stuck were the ones that reduced repetitive cognitive and administrative burden without disrupting workflow: documentation support, coding optimization, chart review, and specialty-specific clinical summarization. Dr. Sadeghian still gets calls from colleagues in California who built these tools with him two and a half to three years ago and are still using them daily.

[00:18:26] Why some clinical GPTs didn't survive — tools that required double documentation or couldn't integrate with the EMR created extra friction rather than eliminating it. When clinicians had to copy-paste outputs or email themselves content to paste into the record, adoption fell off. The lesson: friction removal has to be complete, not partial.

[00:19:56] How Dr. Sadeghian monitors GPT quality over time across 45+ tools — his team relies on active provider feedback, updates specialty guidelines annually (e.g., GINA for asthma and COPD), and flags outputs that are technically correct but regionally inconsistent. He reminds clinicians that AI tools are a second opinion, not a replacement for clinical judgment — especially when regional practice variation means a recommended management approach differs by institution.

[00:21:56] The principle behind "cool technology to enable warm hands" — Dr. Sadeghian's design test for any AI tool is whether it operates invisibly in the background. If it gives clinicians more time to think, listen, and decide, it's working. If it increases clicks, distraction, or documentation theater, it's failing the mission. He applies the same test to vendors: "How can you operate in such a way that nobody notices you?"

[00:23:56] What a patient-facing pediatric chatbot pilot revealed about AI in care access — Dr. Sadeghian built a chatbot for "mommy calls" that answered at a fourth-grade reading level, triaged appropriately (e.g., flagging fever in a two-month-old as a potential sepsis concern), and even deflected off-topic questions correctly. The pilot proved the clinical logic was sound — but the token costs made it unsustainable at scale without a viable funding model. He also sees pre-visit AI intake — where a chatbot gathers the chief complaint and basic history before the physician enters the room — as a near-term direction for health systems.

[00:26:56] Why patient-facing AI creates a cost and liability paradox health systems haven't solved — when patients use AI-driven care guidance for free, they may not come in, eliminating revenue for the health system while the system absorbs the token cost. Liability stays unresolved.

[00:31:26] Where the national AI conversation gets it wrong — Dr. Sadeghian's view from HIMSS and AAP engagement: the field is too optimistic about what models can do autonomously and not ambitious enough about redesigning workflows around them. "Most failures aren't model failures. They are governance and operational failures, and the real opportunity is not better demos, it's better systems."

[00:32:56] Why Dr. Sadeghian is skeptical of outsourcing AI transformation to consulting firms — he sees organizations handing strategic AI direction to firms like Deloitte and Accenture without building internal ownership. His concern: consultants gather stakeholder input, generate a 200-page report, and leave — with no one accountable for execution. He argues that AI transformation requires system-wide ownership across the full workflow, not a report that gets handed off and shelved.

[00:36:26] How early Copilot failures created a lasting confidence barrier for health system executives — many C-suite leaders tried Copilot when it launched as part of Microsoft 365, got poor results, and wrote off AI broadly. Dr. Sadeghian now encounters CFOs and executives who won't revisit the category even as the tools have dramatically improved. He sees this as one of the most underappreciated adoption barriers in healthcare leadership today.

[00:41:26] What drove Dr. Sadeghian to write books rather than just implement — across multiple CMIO roles, he kept seeing the same operational and leadership patterns repeat at different institutions. Writing became a way to move the conversation beyond implementation at any one organization: "I wanna translate what we were learning into something others could apply, challenge, and build on."

‍

‍Fast 5 Lightning Round:

1. What is your favorite book or book you’ve gifted the most?
   Future Shock by Alvin Toffler
2. If you could instantly master any skill, what would it be?
   Piano
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 we should have spent less time documenting care and more time auditing whether documentation was ever the right proxy of care in the first place."
5. What is the last movie or TV show you saw, and what did you think of it?
   The Bear. "I think it's one of the most accurate depictions of leadership I've seen because of how clearly it shows how a system fails when communication does."

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