1. insights
  2. AI Catalyst
  3. health technology
  4. ai catalyst pulse cleveland clinics unconventional ai leadership move
Newsletter | AI-Catalyst

AI Catalyst Pulse: Cleveland Clinic's unconventional AI leadership move

Related Topics:
Health Technology
Relevant Roles:
Chief Information OfficersChief Operation Officers

Cleveland Clinic hired an out-of-industry CAIO. How unusual is that, really?

What happened: Cleveland Clinic announced last week that it has appointed Ben Shahshahani as its first Chief AI Officer (CAIO). The appointment struck us as unconventional for a few reasons:

  • Shahshahani comes from outside healthcare, having previously led machine learning at SiriusXM.

  • He has a Ph.D. in electrical engineering, rather than training in data science or medicine.

But is Cleveland Clinic’s choice really so unusual? To find out, we assembled a data set of all the CAIOs we could identify at U.S. health systems. Because the role is so new, this is still a small, exclusive group – only about 15 so far – but the data revealed emerging tensions over the role’s responsibilities.

Why it matters: Our data shows health systems navigating three fundamental tensions in choosing their CAIOs: insider vs. outsider, clinical vs. technical, and “AI-only” vs. “AI-plus.”

  1. Insider vs. Outsider: 92% of CAIOs in our sample have prior healthcare experience – and half previously worked at the same health system.

Cleveland Clinic's choice of an industry outsider is a definite departure from the norm. But health systems are evenly split between hiring an “insider” from their own ranks versus an “outsider” who worked elsewhere in healthcare.

Consider the contrasting approaches of two California academic medical centers. UCSF tapped Sara Murray, a physician who not only has worked there for a decade – she even got her medical degree at UCSF. UC Davis Health, by contrast, hired Dennis Chorensky, whose background includes roles at The White House and UnitedHealth Group but not in health systems.

The insider-outsider choice signals your priorities: Do you value healthcare expertise, existing relationships, and fluency in your internal systems? Or would you rather import fresh perspectives?

  1. Clinical vs. Technical: 42% of CAIOs in our sample have clinical training (MD or equivalent).

Should AI be led by those who've practiced medicine or those with deep technical expertise?

A clinically trained CAIO, like Ainsley MacLean at Kaiser Permanente (who serves as both CMIO and CAIO), might more easily gain the trust of skeptical physicians and intuitively grasp patient care implications. But a technical expert like Joe Depa, the Chief Data and AI Officer at Emory who previously served as Accenture’s global AI lead, might better push the boundaries of what's possible, including in non-clinical use cases.

Your choice here speaks to how you view AI's role: Is it primarily a tool to enhance existing clinical processes, or a transformative force that might radically reshape your organization?

  1. “AI-Only” vs. “AI-Plus”: 58% of CAIOs also have non-AI responsibilities in their title.

Is your Chief AI Officer just a “Chief AI Officer”? Or do they have other responsibilities – such as, say, a “Chief Data, Analytics, and AI Officer”? This choice reflects your view on AI's place in the organization: Is it a specialized function deserving dedicated leadership or an integral part of a broader digital transformation strategy?

So, to return to our original question, how unusual was Cleveland Clinic’s move? On the whole, they’ve chosen the road less traveled by making an AI-only, non-clinician, out-of-industry hire. They seem to be prioritizing radical innovation over incremental improvement, technical expertise over clinical intuition, and specialized AI leadership over integrated digital strategy.

As you consider your own CAIO strategy, remember that the right choice depends on your health system's AI maturity, culture, and strategic priorities. Our analysis suggests many health systems are playing it safe with insider clinicians in combined roles. While this approach has merits, it might limit AI's transformative potential.

Questions to consider:

  1. Which of these three tensions – insider vs. outsider, clinical vs. technical, and AI-only vs. “AI-plus” – feels most critical for your organization to resolve right now?

  2. How might you structure your AI governance to maximize the strengths and mitigate the weaknesses of your chosen CAIO profile?

  3. If you were to make an unconventional CAIO choice (as Cleveland Clinic did), what specific steps would you take to ensure its success?

From 90 minutes to 9: How Emory's AI gambit is reshaping its appeals writing for rev cycle

What happened: Emory Healthcare is leveraging AI to transform its revenue cycle management. They've implemented DoximityGPT, a HIPAA-compliant tool using GPT-4, to draft appeals letters, patient education protocols, and letters of patient support within seconds.

Early results are striking. Dr. Carla Haack, Emory’s Chief Financial Informatics Officer, says that, in her personal experience, the time to write a letter of patient support has fallen from 90 minutes to 9 minutes, while more than 75% of AI-drafted appeals are leading to overturned claim denials. Perhaps most importantly, more appeals are being submitted, including some that previously wouldn't have been written at all due to time constraints.

Why it matters: To unpack Emory's experience, AI Catalyst spoke with Dr. Haack. We uncovered several key lessons to consider as you implement AI throughout your own revenue cycle operations:

  1. AI enables “impossible” tasks, not just efficiency gains:

    By dramatically reducing the time required to draft appeals, Emory is now submitting appeals that previously wouldn't have been written at all. As Dr. Haack told us, "Prior to Doximity, there were many denial letters that weren't even getting written because the time-suck for clinicians was too great and the ask from revenue cycle employees was daunting. Now those letters are being written because Doximity is helping to close the gap between clinicians and the drafting process."

  2. Non-physicians may (also) be your AI champions:

    Contrary to expectations, nurses at Emory are using AI tools more extensively than some physicians. This challenges assumptions about where AI adoption will take root and suggests that a broader rollout strategy might uncover unexpected use cases and champions.

  3. AI redefines clinical roles as “expert validators”:

    Rather than drafting appeals from scratch, clinicians now primarily review and customize AI-generated content. Dr. Haack emphasized, “DoximityGPT doesn't replace their expertise; rather, it transforms their roles to expert validators of a pre-drafted letter, rather than clinicians doing an administrative task from scratch.” This shift could fundamentally change how we integrate AI into clinical workflows and job descriptions.

  4. AI bridges expertise gaps faster than traditional training:

    By enabling less experienced staff to draft expert-level content, AI tools could dramatically accelerate knowledge transfer. Dr. Haack noted, "Revenue cycle staff members, even those with limited tenure/experience, are able to identify the need for clinician-supplied documentation and generate a draft of that letter using DoximityGPT."

  5. AI might be key to clinician satisfaction:

    By reducing the administrative burden on tasks clinicians particularly dislike, AI could significantly improve job satisfaction. As Dr. Haack put it, “When it came to appeals writing, medical necessity letters, or ADRs, our clinicians were traditionally starting from scratch which can be time consuming and is a major dissatisfier. Frankly, it was a suck of time they didn’t have.”

Questions to consider:

  1. What tasks in your organization are currently “impossible” due to time constraints that AI might make feasible?

  2. How might broadening access to AI tools in your organization uncover unexpected use cases or champions?

  3. How could reframing clinicians as “expert validators” of AI-generated content impact your approach to clinical workflow design and job satisfaction?

Why your team's AI outlook may be more promising than you think

What happened: The past few weeks have seen the publication of several new surveys on clinician attitudes toward AI – and some of their findings are quite positive:

  • A Health Foundation survey

    found that 76% of NHS staff support using AI for patient care, and 81% favor its use for administrative tasks.

  • An internal survey at HCA Healthcare

    found that 89% of nurses testing AI technology for nurse handoffs found it "somewhat" or "very" helpful.

  • More anecdotally, OhioHealth’s COO told Healthcare IT News that their nurses “are fully embracing the AI tools we are bringing in to better support them in their day-to-day administrative work.”

Why it matters: These positive sentiments are, perhaps, surprising given the backdrop of past clinician skepticism and organized protests against AI in healthcare. But they underscore that clinician attitudes towards AI are diverse, nuanced, and likely changing fast. That means you have a window of opportunity to help set the tone for your organization's AI adoption.

We're currently researching how some health systems are doing this effectively (please let us know about your own experiences!), but here are our early takeaways:

  1. Leadership’s silence on AI isn't protecting staff; it's leaving them in the dark.

    While many executives are hesitant to discuss AI initiatives prematurely, this caution may be counterproductive. Our early research suggests a significant communication gap between leadership and frontline staff about future healthcare technologies. Remember: Even if you're tired of talking about AI, your staff may still know little about it … and in absence of knowledge, anxiety and resistance may fester.

  2. At many organizations, middle managers – who often are key players shaping staff sentiment – are being bypassed.

    Our data suggests, for instance, that nurses are most likely to trust their immediate managers for updates on nursing news and staffing. Yet these middle managers are often left out of the AI conversation. If you neglect to arm these key personnel with AI knowledge, you’ll miss an important opportunity to shape staff perceptions.

  3. To win your staff’s AI support, “concrete” beats “abstract.”

    While staff may express general concerns about AI, they often show significant enthusiasm for specific applications, particularly administrative ones. As one physician leader noted to us, "When you start to ask about specific tasks, you see significant enthusiasm." This suggests that framing AI in terms of tangible benefits rather than abstract concepts could increase acceptance.

  4. Don’t assume that younger generations will be your AI champions (or that older generations will be your AI skeptics).

    Contrary to expectations, we’ve seen that younger staff familiar with consumer AI tools like ChatGPT aren't automatically championing healthcare AI. Their familiarity might actually breed skepticism, while experienced staff seeking efficiency could be surprisingly receptive to AI solutions.

Questions to consider:

  1. How can we share AI developments with staff, even if we're not ready to implement specific AI tools, to build trust and gather valuable input?

  2. What would an "AI Ambassador" program for middle managers look like in our organization, and how could it improve our AI communication and adoption strategies?

  3. How can we create regular opportunities for frontline staff to provide input on potential AI applications, leveraging their expertise and potentially increasing buy-in?

AI Strategy Quick Hits

Noteworthy moves from peers to implement AI technologies

Health systems nationwide are continuing to roll out AI-powered clinical documentation technologies. Read our briefing for more information on how to implement these tools effectively:

This week also brought news of several attempts to more deeply integrate AI into nursing workflows:

New perspectives on AI strategy from executives at U.S. health systems:

Additional strategy quick hits:

Emerging Use Cases

New capabilities that indicate AI’s potential

A new study from NYU Langone Health reveals that AI is already, in some important respects, a better communicator than most doctors.

The study used Generated Draft Replies, an EHR-integrated tool powered by GPT-4 through Epic. Comparing 344 message-response pairs, researchers found AI drafts matched or outperformed human responses on key metrics. Physicians rated AI similarly on information quality and usability, while AI significantly outscored humans on communication style (3.70 vs 3.38 on a 5-point scale). Perhaps most strikingly, AI responses were perceived as much more empathetic.

However, the study also revealed the tool's limitations: AI responses were 38% longer and written at a higher grade level than human-written responses, potentially making them less readable for patients.

Research opportunity: AI Catalyst just launched a research sprint on patient messaging use cases. If you've implemented or piloted these tools, please contact Daniel Kim to contribute to our research.

Other emerging use cases:

Market moves

A round-up of AI company announcements and stories

Policy Updates

Understanding the evolving AI regulatory and legislative landscape

The big news this week: Leaders of the bipartisan Senate AI Caucus are set to introduce legislation to significantly expand AI reimbursement in healthcare. Expected after the August recess, the bill would direct CMS to reimburse AI-powered medical devices under Medicare.

This initial focus on medical devices reflects where policymakers have the most data and confidence in structuring reimbursement. It also addresses a major reimbursement gap: Of nearly 900 FDA-approved AI-enabled medical devices, only a handful currently have CMS codes for Medicare reimbursement.

The bipartisan support is an encouraging sign for the bill’s prospects, but don’t expect fast changes. Even if the bill passes this year, implementation likely won’t happen until fiscal year 2026, as CMS will need time to develop and incorporate new payment rules.

Other emerging stories on AI policy:

The News in Numbers

An interesting data point that caught our eye

Predicted growth rate of AI in healthcare market between 2024 to 2032. The market size was 6.1 billion USD in 2023 and is projected to reach 57.2 billion USD by 2032.

Expert Insights

For further reading, articles, videos, and podcasts that we found insightful

Upcoming AI Catalyst Events

  • ‘Candid Conversations’ with leading AI vendors:

    These intimate roundtables offer the chance to speak frankly with vendors about their AI tools, learn about their future roadmaps, and talk with your peers about implementation challenges and real-world outcomes. Sessions include a live demo, Q&A with vendor executives and sometimes an end user, and time to debrief privately with your health system peers. If you’re already using a vendor’s products, we encourage you to attend to share your experiences. If not, this is a great opportunity to learn about new tools on the frontiers of AI. To support a frank dialogue, these events will not be recorded.

    • August 7: Suki’s AI-Powered Documentation Solutions

      • Featuring Suki’s Punit Soni, CEO and Founder and end user Bobby Dupre, MD CMIO, Franciscan Missionaries of Our Lady Health System. Click to register.

    • August 15: Augmedix's AI-Powered Documentation Solutions

      • Featuring Augmedix's Dr. Alex Stinard, CAIO, and Ian Shakil, Founder and CSO. Click to register.

    • August 21: Qventus’ AI-Powered Patient Flow Solutions

      • Featuring Qventus’ CEO Mudit Garg and end user Kim Post from HonorHealth. Click to register.

    • August 28: Ambience’s AI-Powered Documentation Solutions

    • September 4: LeanTaaS’ AI-Powered Patient Flow Solutions

      • Featuring LeanTaaS’ Founder/CEO Mohan Giridharadas and VP of Client Services Jason Harber. Click to register.

  • September 11: AI Clinical Documentation: What's Now – and What's Next?

    • Discover how early-adopter health systems are transforming clinical documentation with AI and gain actionable insights from top executives and real-world case studies to ensure you’re making the right choices for your organization. We’ll be joined by executives Doug Gentile, SVP, Information Technology, UVM; Lance Owens, CMIO, UMH-West; and Todd Richwine, CMIO, THPG-THR. Click to register.

  • September 25th: Deep Dive: How AI is Transforming the Revenue Cycle

    • Explore how emerging AI technologies are reshaping revenue cycle management, hear from early adopters about their experiences and lessons learned, and consider the potential impact to your bottom line. Click to register.