Will AI in healthcare lead to consolidation among tech providers?

There's a potential for consolidation, with Epic seen as a dominant 'system of record'. Companies offering single-point solutions may struggle unless they evolve into platforms that address multiple needs.

Is the patient, rather than the physician, the primary 'human in the loop' for healthcare AI?

The speaker argues that patients may increasingly be the 'human in the loop' due to accessibility and trust issues with traditional healthcare. This shifts the focus for developing patient-facing AI tools.

What are the risks of patients using AI for health queries?

Patients may lack the expertise to discern accurate AI responses from dangerous ones. The tools for consumers might need to be more 'doctor-like' to ensure safety, as they currently lack the nuanced diagnostic capabilities of human physicians.

Why are hospitals adopting AI cautiously, one tool at a time?

Hospitals are fearful of making mistakes and face complex approval processes. They are adopting a 'dipping toes in the water' approach, starting with less critical tools like scribes before moving to more integrated platforms.

What is the historical precedent for technological adoption in healthcare?

The adoption of hospitalists 30 years ago faced similar resistance. While AI adoption might seem slow, it's considered 'rocket pace' compared to the average 17 years for new healthcare advances to be implemented.

How do regulations like HIPAA affect AI integration in EHRs?

Embedding advanced AI tools directly into EHRs raises liability and regulatory concerns (HIPAA, ONC). This leads to a 'thousand flowers blooming' scenario with organic, external use of AI tools before institutional integration.

Key Moments

AI in Healthcare: Why Hospitals Are Moving Cautiously Toward Consolidation with Bob Wachter, MD

Stanford Online

Education6 min read37 min video

Mar 30, 2026|1,097 views|59|2

Stanford Stanford Online

Save to Pod

Key Moments

TL;DR

AI in healthcare faces a slow adoption due to integration complexities and fear of failure, despite its magic in daily life. The real battle is between Epic as the incumbent and other players, not just AI model supremacy.

Key Insights

Electronic Health Records (EHRs), particularly Epic, hold a massive incumbency advantage and are likely to remain the core platform for healthcare data and patient portals.

Consumer use of AI for health queries is surging, especially in 'care deserts,' highlighting patient demand and empowerment, but posing risks due to users' limitations in discerning accurate information.

The adoption of AI tools in healthcare is proceeding cautiously, starting with less critical applications like digital scribes and chart summarization, to build trust and avoid high-profile failures.

The future of AI integration in healthcare might see consolidation on the enterprise side, focusing on platforms rather than numerous point solutions, to manage costs, integration, and security risks.

While AI tools are magic in daily life, their integration into complex healthcare systems requires understanding motivations, payment structures, and stakeholder dynamics, necessitating a slower, foundational approach.

The average time for implementing new healthcare advances is 17 years, making the current pace of AI adoption, spurred by tools like GPT and Gemini, appear as 'rocket pace' by comparison.

The enduring dominance of EHRs like Epic

The conversation on AI in healthcare frequently circles back to the central role of Electronic Health Records (EHRs), with Epic identified as a powerful incumbent. Despite the rapid advancements in AI models, many believe EHRs will serve as the foundational platform for healthcare data, including patient portals. This 'massive incumbency advantage' means new AI tools and organizations must overcome a significant barrier to become the platform of choice. While it's tempting to be skeptical of an EHR company's ability to adapt to AI, Epic's success with EHRs suggests they might also navigate the AI landscape effectively. This creates a scenario where third-party tools might serve as 'ornaments on a Christmas tree' attached to Epic, offering capabilities generally good but perhaps not as cutting-edge as specialized vendors, yet providing the benefits of integration, reliability, and vendor perpetuity. The integration of AI into the EHR is seen as crucial, moving data from various sources into a unified system that can be queried by both professionals and patients.

The rise of the patient as the human in the loop

A significant shift discussed is the potential for patients, rather than solely physicians, to become the 'human in the loop' for AI technologies. This is driven by increased patient access to AI tools like GPT and Gemini for health queries, particularly evident in 'care deserts' where access to traditional healthcare is limited. The democratization of care, amplified by skepticism towards professional expertise and institutions, turbocharges this trend. While these tools provide valuable information, a critical concern is the patient's ability to discern accurate AI-generated advice from potentially dangerous misinformation. Unlike physicians who possess subtle, learned expertise that informs their AI queries, patients may struggle to formulate complex prompts or critically evaluate responses. The implication is that consumer-facing AI tools need to be designed with a more 'doctor-like' approach, offering guided interactions rather than direct, unfiltered information, to ensure patient safety amidst this evolving dynamic.

The cautious, staged approach to AI adoption

Unlike the rapid adoption seen in other sectors, healthcare is taking a decidedly cautious approach to AI implementation. The industry is progressively 'dipping its toes in the water,' starting with tools like digital scribes and chart summarization. This measured strategy is intended to build trust, gain buy-in from clinicians, and avoid catastrophic, high-profile failures that could derail the entire field, mirroring past setbacks like the initial challenges with AI in sepsis diagnosis. The fear of making a mistake that could harm patients or lead to significant media backlash is a driving force. This staged implementation allows for learning and validation, gradually moving towards more complex applications like prior authorization assistance. The process involves gaining acceptance from healthcare professionals, ensuring the tools enhance rather than hinder workflow, and building a solid foundation before attempting more ambitious, potentially higher-risk use cases.

The battle between Epic and the broader AI ecosystem

The central conflict shaping AI in healthcare isn't solely about which AI model (like OpenAI's or Google's) will dominate, but rather the strategic positioning of Epic against a host of other players. The question is whether Epic will remain the core 'system of record,' integrating AI capabilities as features, or if external vendors, including large tech companies and startups, will vie for dominance. This dynamic is further complicated by the existential threat AI poses to traditional software vendors, as LLMs could abstract away the need for many specialized, paid-for tools. The industry is at a crossroads: Epic could evolve into the integrated AI platform, or a more fragmented ecosystem could emerge. However, the practicality for large institutions like UCSF and Stanford suggests a need for consolidation, with room for perhaps only a few essential platforms that solve multiple problems, rather than dozens of niche point solutions.

The role of 'shadow IT' as an organic adoption mechanism

The phenomenon of 'shadow IT,' where individuals or departments use technology solutions without explicit IT department approval, is viewed by some as a potentially positive force in healthcare AI adoption. This organic use, often driven by clinicians seeking immediate value, can act as a pressure release for the slow-moving enterprise-level implementation processes. It allows for experimentation and adoption of AI tools, like advanced clinical insight engines, outside the strictures of institutional approvals and extensive training modules. While this approach carries risks, such as potential HIPAA violations if protected data is mishandled, it also allows for individual and system-level learning. It highlights that users are already leveraging these tools, creating a demand that institutions may eventually need to address by integrating them officially, ensuring compliance and security, and thereby driving further, more regulated adoption.

Consolidation as a likely future for AI solutions

The current model of hospitals acquiring numerous individual AI solutions for different needs—ambient documentation, revenue cycle management, scheduling—is seen as unsustainable and inefficient. The high cost of integration, security vulnerabilities, and the burden on users to navigate multiple platforms are significant drawbacks. Experts predict a substantial consolidation wave on the enterprise side, as healthcare systems realize the benefits of integrated platforms. This mirrors historical trends where fragmented EHR markets eventually consolidated. The drive towards fewer, more comprehensive solutions will likely favor vendors capable of offering a suite of services rather than single-point solutions. This consolidation is partly driven by the desire to streamline workflows, reduce IT overhead, and enhance security, ultimately making the adoption of AI tools more manageable and cost-effective for larger healthcare organizations.

The inevitability of AI adoption, despite resistance

The history of technological change in healthcare, exemplified by the slow but eventual widespread adoption of hospitalists and EHRs, suggests that AI will follow a similar path. While initial resistance from professionals concerned about patient care, job security, or disruption to established practices is expected, many will eventually need to adapt. The analogy of electronic health records is particularly relevant; just as physicians who resisted EHRs eventually had to adopt them to practice, similar pressures may arise for AI tools. As certain AI applications, like ambient scribes, become indispensable for efficient patient care, clinicians unwilling or unable to use them may find it increasingly difficult to practice. This gradual but firm shift towards AI integration is seen as an inevitable consequence of technological advancement and evolving professional standards within the healthcare industry.

Mentioned in This Episode

●Software & Apps

●Companies

●Organizations

●Books

●People Referenced

Common Questions

Current AI applications in healthcare include digital scribes and tools like Open Evidence for clinical insights. While broader AI adoption is cautious, these tools are seeing significant implementation in institutions.

Topics

Ai-Ethics Health & Longevity AI & Machine Learning Technology & Innovation AI Adoption Healthcare AI Digital Health EHR Systems Patient Engagement Clinical Decision Support

Mentioned in this video

People

Seth Hayne

Previous guest on the podcast who discussed Epic.

Bob Wachter

Chair of Medicine at UCSF, author of 'The Digital Doctor' and 'The Giant Leap', discusses AI in healthcare.

Magic Johnson

Player on the opposing team (Michigan State) that beat Bob Wachter's University of Pennsylvania team in the 1979 Final Four.

Peter Lee

Mentioned as someone who articulated the difference between expert and novice use of AI in healthcare, contributing to Bob Wachter's book.

Jeff Fowler

Digital reporter for The Washington Post who co-authored a story with Bob Wachter on GPT prompts from patients.

Graeme Walker

An AI-focused doctor who wrote about a patient's trust in a chatbot over professional advice.

Organizations

UCSF

University of California, San Francisco, where Bob Wachter is Chair of Medicine. Mentioned in the context of AI adoption and challenges.

University of Pennsylvania

University where Bob Wachter was the mascot, the 'penquaker', in 1979.

Princeton

An Ivy League university whose basketball team, with Bill Bradley, previously went to the Final Four.

Open Evidence

A tool or platform used for clinical insights, mentioned as being adopted in healthcare institutions.

The Washington Post

Newspaper where Jeff Fowler works, which published a story with Bob Wachter on patient use of GPT.

ONC

Office of the National Coordinator for Health Information Technology, regulations mentioned in relation to embedding AI tools within EHRs.

Books

The Digital Doctor

Bob Wachter's first book, which discusses the digital transformation in healthcare experienced in the 2000s.

The Giant Leap

Bob Wachter's new book that discusses how AI has transformed healthcare and the missteps in the process.

Software & Apps

Epic

A major electronic health record (EHR) system provider in healthcare, discussed as a potential platform for AI integration and an incumbent advantage.

GPT

A large language model discussed for its capabilities in healthcare queries and potential use by patients and professionals.

Gemini

Google's AI model, mentioned as a tool used by Bob Wachter for various tasks, including pictures, videos, and writing.

Claude

An AI model mentioned by Bob Wachter for its 'charming personality' and used for writing assistance.

Claude for Healthcare

A healthcare-specific version or application of the Claude AI model, mentioned in recent announcements.

ChatGPT for Healthcare

A healthcare-specific version or application of ChatGPT, mentioned in recent announcements.

LLMs

Large Language Models, discussed in the context of consumers using them for health queries.

Companies

Google

A tech company whose integrated services (Gmail, Docs) are compared to Epic's potential role as a unified platform in healthcare.

OpenAI

A prominent AI research lab, mentioned as a potential destination for healthcare data outside of EHRs.

Apple

A technology company, mentioned as a potential destination for healthcare data outside of EHRs.

Locations

Palo Alto

City mentioned in the context of difficulty finding primary care physicians.

San Francisco

City mentioned as an example where finding a primary care doctor is difficult, despite not being a 'care desert'.

Legislation & Policy

HIPAA

Health Insurance Portability and Accountability Act, regulations mentioned in relation to embedding AI tools within EHRs and potential liability risks.