In the age of AI, the economics of knowledge are being rewritten — and nowhere more profoundly than in healthcare. 

Leading healthcare provider Mayo Clinic’s digital transformation over the recent past offers one blueprint for what this revolution looks like in practice. Beginning in 2020, Mayo embarked on an ambitious strategy to create a much deeper and more facile digital capability — including AI.  To prepare the organization for this transformation, the senior leadership at the venerable institution knew they would have to change the way they managed their digital assets.  

In order to accomplish this ambitious goal, Mayo carved out an effort they named The Center for Digital Health. The governance of this group was outside the traditional IT organization, and the center took over a significant part of the IT budget and raised outside funding to help create a new platform for the hospital of the future. A senior executive team, including newly recruited IT talent, was given the task to unify, govern, and operationalize Mayo’s data at scale.

This was a difficult task. Traditionally, medical knowledge has been scarce, slow to move, and deeply siloed. Expertise was constrained by geography, credentialing, and legacy systems. Mayo was no exception. Data resided in fragmented on-premise systems, sometimes even on thumb drives. Heroic efforts were required to assemble data for traditional AI. Even the Epic EHR, while modern, functioned as a closed silo.

Within the Center for Digital Health, Mayo implemented a decentralized data governance model, enlisted data stewards across the organization, and migrated data to the Google Cloud Platform (GCP), with ERP data moved to Oracle Cloud. Most critically, they built a Longitudinal Patient Record database in a new data system that allowed for different types of data to be accessible in one platform. (The tool was built to Fast Healthcare Interoperability Resources Release 4 standards and the tool was BigQuery from Google.)

This resource created a unified, indexed, and context-rich store of structured and unstructured medical data — including images, waveforms, notes, genetic data, and more — all within a HITRUST-certified virtual private cloud. This foundational shift made data not only accessible but operable at scale.

The data foundation enabled a powerful wave of AI adoption, from predictive models to generative tools, reshaping clinical care, research, and regulation. Furthermore, this investment in transforming the data resource unleashed what Mayo now experiences as “knowledge liquidity,” a state where information flows freely across research, diagnostics, policy, and patient care. It’s a foundational enabler of Mayo’s AI strategy and a model for science-based organizations everywhere.

In addition to the technological investment, Mayo delivered extensive training on the new tools and techniques for their administrative and healthcare staff, and provided a centralized funding system where doctors, researchers, and departments could ask for $25,000 to $250,000 in seed funding for projects. These initiatives were overseen by a steering committee of senior executives at Mayo.

This innovation budget enabled many groups to create practical prototypes. The approach gave birth to many projects across the organization: from superior knowledge access to streamlining administration, to new treatment algorithms.

The rise of commodity knowledge: Universally available expertise

The first shift was the radical commoditization of baseline medical knowledge. Thanks to LLMs and intelligent search, vast repositories like PubMed were now accessible to anyone in easier-to-use natural language queries. Moreover, the GenAI tools made complex clinical literature vastly more usable as they can provide explanations, summaries, and synthesize recommendations to any level of expertise or domain of knowledge. This democratization empowers healthcare workers across the spectrum.

Administrative and operational improvement

Ambient GenAI systems like Abridge were being deployed to transcribe and summarize clinician-patient interactions in real time. Initially designed for primary care, Abridge required retraining for Mayo’s specialty needs, highlighting that while AI can generalize, broad implementation in practical settings often requires some localizing for best use.

With the data platform and generative AI (GenAI), Mayo began transforming administrative functions. The first major use case was indexing and vectorizing internal policy and admin documents for instant semantic search.  This reduced friction and eliminated outdated content.

Success in this domain expanded to nursing workflows (automated flowcharting), documentation, billing, and intranet knowledge management. These are not glamorous, but they produce immediate value.

Strategic AI: From predictive models to research acceleration

The ability to instantly access credible, current information shifted the locus of competitive advantage. No longer was it about owning knowledge, but about applying and personalizing it rapidly. Mayo is finding that as AI levels the playing field for commodity knowledge, the new advantage lies in context-sensitive application and institutional tuning.

For example, Mayo is leveraging AI to accelerate discovery and implement insights. Once the data estate was governed and unified, Mayo deployed tools to develop predictive models at scale.

Innovations include:

• AI-ECG for LVSD: A model to detect Left Ventricular Systolic Dysfunction (LVSD) using subtle waveform anomalies in ECGs, often missed by human eyes.
• HFpEF Detection: An FDA-approved algorithm, developed with Ultromics, to identify Heart Failure with Preserved Ejection Fraction (HFpEF) via standard echocardiograms.

These tools are in live clinical use at Mayo. The institution has chosen to implement them internally before publication, creating a form of knowledge arbitrage — gaining performance benefits before broader diffusion. This changes the competitive dynamics in the care provision space by shrinking the time from discovery to application.  Institutions must race not just to discover, but to operationalize knowledge faster than rivals.

Translational infrastructure: Governance, compliance, and the role of T-Rex

Accelerated discovery also requires responsible deployment. Mayo’s T-Rex system (Translational Regulatory Expert System), co-developed with the FDA, provides this critical translational infrastructure. Originally launched as a rules-based tool, T-Rex determines whether a new AI model constitutes Software as a Medical Device (SaMD) and requires FDA review. If so, it auto-generates submission materials, cutting approval timelines from 4–5 years to under 18 months.

More recently, T-Rex was enhanced with GenAI, making it a smarter workflow tool that learns and adapts. AI tools like T-Rex enable safe, compliant, accelerated translation of research into practice.

GenAI and the Digital Twin Frontier

Mayo’s most ambitious GenAI initiative is Maia, a clinician-facing chatbot capable of real-time conversation powered by Retrieval-Augmented Generation (RAG). Behind Maia is the fully vectorized longitudinal patient record and the body of internal and external medical knowledge. This allows Maia to generate differentials, identify like-patients, and suggest treatment options based on outcomes of “digital twins.”

Statistically, Maia has exceeded physician-level accuracy in differential diagnosis — though such claims remain controversial. Due to its capabilities, Maia is still in limited pilot use due to policy, risk, and security debates.  Tools like Maia preview a future where AI augments every clinical decision — not as a replacement for physicians, but as an ever-present intelligence layer.

Looking to the future: Building on solid foundations

In healthcare, the creation of a productive data foundation is critical, and this may necessitate new organization and funding structures to unlock this sea of data in a reasonable time frame. There were four pillars to that foundation:

• A new IT organization, funding, and governance to enable a new age platform to increase functionality and unlock access.
• Deep and broad staff training, cloud access, and tools to enable application and scale.
• Investment in innovations that apply new AI and GenAI capabilities and methods to solve problems that had not been solved before.
• A facile combination of GenAI tools and traditional regulation to help manage the complexities of governance, risk, and compliance to ensure safe and responsible use.

A new perspective: Co-evolution of AI, science, and care

Mayo’s team is now half a decade into reinventing its use of human and machine knowledge. They have created technological, organizational, and funding structures to unleash the combination of their brilliant talent, the vast data they have, and tools to make sense of that data. The thoughtful observer will see that Mayo Clinic’s transformation is not just a case study in AI. It is a model for strategic reinvention.

They are creating an organizational system that co-evolves in a high-risk, high-impact domain by integrating machine intelligence, regulation, policy, and oversight in a way that can move science forward while creating an ongoing competitive advantage for one of the great healthcare institutions of the world.

The good news is that they can evolve faster. The bad news is that just about every science-based business will need to. Mayo has an approach that many can learn from – because this phenomenon of faster co-evolution is occurring in every knowledge-based business in the world at an ever faster pace.

About the Authors:

Ajai Sehgal was Chief Data & Analytics Officer at Mayo Clinic from 2020 through 2024. He was previously CTO/CIO of Hootsuite and senior executive with Expedia for over a decade.

John Sviokla is an executive fellow at Harvard Business School, where he was formerly an associate professor for over a decade. He is co-founder of GAI Insights.  John was previously US Marketing Leader for PwC Advisory and Vice Chairman of Diamond Technologies.  

Randy Bean is the author of Fail Fast, Learn Faster: Lessons in Data-Driven Leadership in an Age of Disruption, Big Data, and AI, and a contributor to Forbes, Harvard Business Review, and MIT Sloan Management Review. He has been an advisor to Fortune 1000 organizations on data & AI leadership for 3+ decades.