The opening keynote of the 2026 HIMSS Global Health Conference and Exhibition on March 10 in Las Vegas brought together three voices with sharply different vantage points on innovation, and the result was a surprisingly coherent message: healthcare’s AI transformation era is no longer arriving — it’s here, and the hard work of making it function inside real organizations has begun.
HotSpot Take
Healthcare AI has moved from theoretical to operational, but three HIMSS26 keynoters agreed: technology alone doesn’t transform care. Process, culture, and data architecture determine whether it sticks.
Hal Wolf: Grounding the AI Conversation
HIMSS President and CEO Hal Wolf used his opening remarks to push back gently on the hype that has surrounded AI in healthcare. He acknowledged that the technology is already producing tangible results in specific operational areas, citing bed utilization, supply chain management, and staff scheduling as domains where AI is being deployed with measurable effect. “If you integrate it and make the decision to bring those analytics forward, it improves efficiency,” Wolf said.
But Wolf was direct about what AI cannot do on its own. Health systems that invest in the technology without simultaneously redesigning the workflows and workforce models it’s meant to support will see those investments fail. “If you’re going to make the investment but not redesign processes to take advantage of it, then it fails because we don’t adjust our people or our processes,” he said.
He offered a longer view on the pace of adoption, noting that only a few years ago, fewer than 5% of organizations were running AI in production environments. More recently, AI capabilities have begun appearing as embedded features within major enterprise healthcare platforms — a shift Wolf described as the beginning, not the arrival, of AI’s clinical role. “We’re just at the start of the operational scope,” he said, particularly with respect to clinical decision support.
“With any new tool that’s introduced, it needs to be objectively looked at by governance before the tool is used. Bad data in, bad information out.” — Hal Wolf, President and CEO, HIMSS
Governance and data quality anchored much of Wolf’s practical guidance. New tools must move through rigorous oversight before deployment, he argued. “With any new tool that’s introduced, it needs to be objectively looked at by governance before the tool is used. Bad data in, bad information out.”
Wolf also highlighted the role of nursing in digital transformation, describing bedside nurses as the closest link between clinical systems and patients and, by extension, as essential contributors to the design of tools that will actually work at the point of care. “The closest relationship between digital health technology and the patient is at the nursing level,” he said.
He closed by framing the ultimate goal of healthcare technology plainly. “Nothing’s easy about healthcare. It’s about educating and putting people in the position to make a better decision the next day. If we do that, we all win.”
Jon McNeill: “The Algorithm” Applied to Healthcare
The second keynote came from Jon McNeill, CEO of DVx Ventures and former President of Tesla and COO of Lyft. McNeill did not come to the HIMSS stage as a healthcare expert. He came as someone who has scaled organizations through environments that resisted change, and he argued that the principles involved transfer directly.
His central framework, which he calls “the algorithm,” is drawn from his time at Tesla, where he helped grow revenue from roughly $2 billion to $20 billion in approximately 30 months. The framework holds that innovation is not a department or a special initiative but a system — one that any organization can operate continuously at the leadership level.
“At Tesla, there’s no center of innovation. There are no innovation teams. The belief is that everybody is innovation, but innovation is a framework. There’s a framework that drives it, and internally that framework is called the algorithm. It’s an always-on system driven by leadership, and you practice this weekly,” McNeill said.
The algorithm begins with simplification. Complexity, McNeill argued, is not the same as necessity. The first step is to question every requirement — aggressively, repeatedly, and without assuming that existing processes exist for good reasons. “You’ve got to question every requirement, even if it’s offensive,” he said. He pointed to the Tesla purchasing process as a concrete example, recounting how a single question he pursued for weeks ultimately revealed an unnecessary step that, once removed, cut cost and friction from the transaction.
The next steps are to delete every process step that can be eliminated, then simplify and optimize what remains, then accelerate the cycle time of what’s left, and only then, as the final step, automate. The sequence matters, McNeill emphasized. “If you do automation first, it slows the process down.”
“Your goal shouldn’t be to improve patient satisfaction. It shouldn’t be to deploy AI. Instead, it should be to reduce diagnosis cycle times by 50%.” — Jon McNeill, CEO, DVx Ventures
He drew a direct line to healthcare by inviting the audience to count the clicks required to complete common clinical or administrative tasks — scheduling a follow-up, ordering medication, submitting a prior authorization. When Tesla simplified its online purchase flow to a handful of steps, according to McNeill, sales grew roughly 20 times. “Your goal shouldn’t be to improve patient satisfaction. It shouldn’t be to deploy AI. Instead, it should be to reduce diagnosis cycle times by 50%. It should be to cut patient throughput time in half,” he said.
Three cultural ingredients, McNeill argued, determine whether the algorithm works in practice: expanding the definition of the product to include the full customer experience from beginning to outcome; injecting urgency and accountability into every initiative; and building the discipline to use the system internally before deploying it to others.
John Halamka: Building the Platform That Makes It Possible
Dr. John Halamka, President of the Mayo Clinic Platform, closed the opening keynote sequence with a different kind of argument. Where Wolf provided perspective, and McNeill provided a process framework, Halamka provided a working example of what happens when a major health system commits to building the data infrastructure that advanced AI actually requires.
“You won’t see the term AI in my slides. I did that purposefully, because having done this for 40 years, what you know is the technologies will change, but the engineering principles will stay the same.” — Dr. John Halamka, President, Mayo Clinic Platform
In a notable departure from the industry’s prevailing vocabulary, Halamka opened by telling the audience they would not see the term “AI” anywhere in his presentation slides. “I did that purposefully,” he explained, “because having done this for 40 years, what you know is the technologies will change, but the engineering principles will stay the same.”
The Mayo Clinic Platform has assembled a dataset of more than 10 million patient records spanning decades, encompassing structured and unstructured data, imaging, genomics, and wearable data. The work required solving a problem most organizations have not yet confronted at this scale: ensuring that every record, regardless of age or format, is effectively de-identified before it can be used to train or validate AI tools. Halamka described this as a fundamental architectural challenge. “It really does require rethinking your architecture. So sure, you’ll have your systems of record, your EHR, your ERP, but how do you wrap those in a mechanism that’s standard-based, scalable, and future-proof?” he said.
Halamka offered a personal account of AI’s clinical potential in his own care. He described receiving a remote cardiac assessment through 14 AI algorithms, using only a consumer wearable device, that evaluated his heart structure, valve function, and circulation without an in-person visit. The result was clinically actionable: a diagnosis that identified a specific conduction abnormality treatable with medication. “We did it. And the answer is, I don’t have any structural issues. I just have a node that conducts poorly,” he said.
His larger argument was that AI-augmented clinical capabilities are advancing to the point where withholding them from patients may itself become an ethical question. The healthcare profession, he suggested, may soon face a standard-of-care reckoning over which AI tools have become too effective not to use.
At the same time, Halamka was careful to ground the conversation in what technology cannot replace. Regardless of the sophistication of the platform underneath, the core of care delivery must remain human, he argued. All of the technology and advanced clinical decision support in the world do not change that fundamental orientation.
A Sequence That Held Together
Conference programming is not always more than the sum of its parts, but this keynote sequence worked. Wolf established that AI is a tool requiring human process and judgment, not a solution on its own. McNeill showed what disciplined, human-led process redesign looks like when it’s applied without sentimentality toward existing workflows. And Halamka demonstrated what becomes possible, clinically and operationally, once the data infrastructure catches up to the ambition. For a healthcare industry still wrestling with the gap between AI’s promise and AI’s practical adoption, it was a grounding way to begin the week.
Photo credit: HIMSS26
— This original article was created with AI support.
