The Lean Algorithm: Why AI in Healthcare is a Process Problem

A conversation with Dr. Gary S. Kaplan, MD—Board of Stewardship Trustees member at CommonSpirit Health, CEO Emeritus of Virginia Mason Franciscan Health, and 2026 Deming Lifetime Achievement Award Recipient —on why automating a bad process only accelerates waste, how AI can revolutionize clinical training, and the enduring power of the human-in-the-loop.

By Louise Servoin · 2026-08-13 · 12 min read

## Gary S. Kaplan, MD – CEO Emeritus, Virginia Mason Franciscan Health; Member, Board of Stewardship Trustees, CommonSpirit Health

An internationally renowned pioneer of Lean healthcare, clinical system integration, and patient safety. Alumnus of the University of Michigan Medical School and Member of the National Academy of Medicine. Based in Seattle, WA, United States.

Few leaders have reshaped the operational DNA of modern healthcare as deeply as Dr. Gary S. Kaplan. When he took the helm of Seattle’s Virginia Mason Health System in 2000, the organization faced severe financial deficits and the entire US healthcare sector was grappling with systemic patient safety failures. Rather than benchmarking against other struggling medical centers, Dr. Kaplan looked to industrial engineering—specifically Boeing’s adaptation of the Toyota Production System.

The resulting framework, the Virginia Mason Production System (VMPS), turned Virginia Mason into a global model of zero-defect clinical quality, earning it "Hospital of the Decade" honors. Recently honored with the prestigious 2026 Deming Lifetime Achievement Award from Columbia Business School, Dr. Kaplan now serves on CommonSpirit Health's Board of Stewardship Trustees, one of 14 members overseeing quality and mission across its 21-state national network.

We sat down with Dr. Kaplan to discuss why today’s generative AI boom mirrors the early days of Lean management, how AI-assisted training can accelerate clinical education, and why the most dangerous thing a health system can do is automate a broken process.

## The Lean Architect of Digital Workflows

Dr. Kaplan, your entire career has been built on dismantling the traditional, physician-centric design of hospitals and replacing it with patient-centered, highly standardized workflows. Today, healthcare is infatuated with artificial intelligence. How does your experience implementing Lean management shape how you view this AI wave?

The parallel is striking. In the early 2000s, healthcare leaders thought they could solve their quality and cost crises by buying expensive new clinical equipment or building bigger facilities. They focused on inputs rather than processes. Today, we see a similar obsession with AI. Organizations are racing to buy the latest algorithms, treating AI as a magical "black box" that will solve their productivity issues.

But as we learned through the Toyota Production System, technology is an enabler, not the driver. If you do not understand the underlying workflow, any tool you introduce will fail.

In Lean, we focus heavily on the seven wastes: time, motion, inventory, processing, defects, transportation, and overproduction. If you take a clinical workflow that is riddled with these wastes and layer an AI tool on top of it, all you succeed in doing is automating your defects. You make bad things happen faster.

Our starting point at Virginia Mason was always to ask: "What is the specific waste or defect we are trying to eliminate?" Only when the process is standardized, clean, and stabilized can you introduce technology to accelerate it. AI in healthcare needs to not just provide a technical solution but help us redesign clinical care.

## AI-Assisted Training: Scaling the "Gemba"

Your educational initiatives through the Virginia Mason Institute have trained thousands of healthcare leaders globally, including organizing hands-on study trips to Japan to study continuous improvement (kaizen). How do you see AI transforming clinical training and professional development?

Training has always been the bottleneck of clinical transformation. When we first introduced VMPS, we faced immense cultural resistance. Physicians operated as autonomous craftsmen; they didn't want standard operating procedures. Overcoming that required deep, highly resource-intensive education. We mandated Lean certification for all senior executives, requiring demonstration of proficiency and intensive classroom and practical sessions.

One of the most exciting frontiers today is AI-assisted training. Traditionally, clinical training is a "one-to-many" lecture model. But adult learning—especially for high-stress roles like critical care or surgery—thrives on personalized, spaced, adaptive feedback.

AI-driven co-pilots have the potential to democratize this expert training. Imagine an AI training system integrated into the clinical workspace that acts as an interactive coach. For example, when we streamlined laparoscopic surgical setups from 74 instruments to 58 to reduce clinical waste, it took weeks of physical workshops to align our teams.

An AI-assisted training platform can instantly personalize how a nurse or resident learns these standardized setups, adapting in real-time to how they process information. It can simulate a Gemba—our physical walk on the shop floor —and present clinicians with standard-work scenarios, challenging their decision-making on the spot. It shifts education from a passive annual credentialing exercise to a personalized, continuous kaizen companion.

However, we must be careful. We cannot allow AI-assisted training to replace actual physical observation at the bedside. You must still go to the Gemba to see where the value is created. But as a personalized training accelerator, AI is the most disruptive tool we’ve seen since the printing press.

"If 10 surgeons do things 10 different ways and there is no value added in those variations, you have a defect-prone system. If you deploy AI to automate that variation, you are simply accelerating your defects." — Dr. Gary S. Kaplan, MD

## Standardization Meets Machine Learning

In your lectures worldwide, you often discuss "Moving from Physician-Driven to Patient-Driven" care. AI developers argue that algorithms can personalize treatment plans for individual patients. Is there a tension between Lean standardization and AI-driven personalization?

There is no tension if you understand what standard work actually is. Standard work is not a straightjacket; it is the foundation for innovation. W. Edwards Deming and Taiichi Ohno taught us that without a standard, there can be no meaningful improvement.

Let’s look at a concrete example from our history: lower back pain. Historically, patients with back pain almost immediately received highly expensive, unnecessary MRIs because doctors practiced in silos. By standardizing the care path—sending most patients directly to physical therapy first—we reduced unnecessary MRIs, dramatically lowered costs, and got patients back to work faster.

AI and machine learning require highly standardized, clean data inputs to function without generating "hallucinations" or biased errors. If your clinical teams do not input data in a standardized way, the algorithm’s output will be noise.

Standardization actually enables AI to personalize care safely. By automating the highly repetitive, low-variance administrative steps—such as clinical documentation, billing prep, or simple triage—we free up the physician's cognitive bandwidth. The algorithm handles the structure, leaving the human clinician to apply judgment and empathy where it matters most.

## The Human-in-the-Loop: Preserving the Heart of Medicine

You worked in your father's hardware store in Detroit as a teenager, where you first learned the value of face-to-face customer service and relationship management. You’ve often expressed concern about "depersonalizing" healthcare through technology. How do we maintain that personal, human connection in an increasingly automated environment?

Those lessons from my father’s store have stayed with me for over sixty years. Healthcare, at its core, will always be about people taking care of people. When I was 14, a local physician who was a regular customer at our store took me on his clinical rounds. That is when I fell in love with medicine—not because of the technology of the operating room, but because of the deep, continuous trust between that doctor and his patients.

If we deploy AI simply to optimize for throughput or transaction speed, we will destroy the essence of our profession. If a doctor spends their entire shift looking at an AI co-pilot dashboard instead of looking into their patient’s eyes, we have failed.

We must design AI systems with a deliberate "human-in-the-loop" architecture. AI should be a decision accelerator, not a decision-maker. It can analyze a million patient records in seconds, but it cannot sit in a hospital room, hold a patient's hand, and understand the cultural or emotional nuance of a family's end-of-life choices.

Our goal with the Virginia Mason Production System was always to eliminate waste so that our clinicians could spend more high-quality, uninterrupted time with their patients. If we use AI to remove administrative waste—the "clerical burden" that drives much of clinical burnout—and give that time back to the patient-physician relationship, then technology will have fulfilled its highest purpose.

## The Golden Thread

We often see a gap between what technology can do and what professionals actually adopt. How do we ensure AI becomes a 'co-pilot' that elevates human judgment rather than a tool that encourages intellectual passivity?

We must continue to emphasize the power of human interaction as part of the healing process. If the critical physician-patient relationship continues to be thought of as essential, then AI tools will be appropriate assets to elevate this relationship and the clinical outcomes generated as key indicators of success. By using technology to aid the elimination of waste, we ensure higher quality technical and functional outcomes, better patient and team member experiences, and we lower the total cost of care. In many ways that’s the holy grail and needs to continue to guide our professional endeavors on behalf of the people we serve.

Tags: AI in Healthcare, Lean Management, Patient Safety, CommonSpirit Health, Clinical Training, W. Edwards Deming, Virginia Mason Institute, Healthcare Innovation