AI is Redefining Healthcare Outcomes by Moving from Silos to Solutions

UCLA Anderson produces transformative leaders who are fluent in the use of the latest technologies. Now in its second year, Anderson’s Healthcare Analytics Symposium (HAS) reflects our commitment to advancing data-driven solutions in complex industries such as healthcare.

Organized and hosted by the Morrison Family Center for Marketing and Data Analytics in collaboration with UCLA Anderson’s Master of Science in Business Analytics program, the 2025 HAS symposium explored how AI tools can be leveraged to connect complex, siloed data and generate valuable insights that enhance healthcare and patient outcomes.

In 2025, HAS drew an even broader, more diverse audience that included students and faculty across UCLA’s departments, Anderson alumni and a strong showing of healthcare professionals and industry leaders. Their discussions focused on innovating healthcare delivery, confronting adoption barriers and bridging research with practice.

The opening symposium highlighted a central tension: While the healthcare industry is awash in data, much of it remains locked in silos created by regulatory complexity, misaligned incentives, legacy systems and fragmentation across providers and platforms. Even the most promising AI and agentic frameworks cannot fulfill their potential without reliable, accessible and diverse data. Is AI the solution? Maybe. But its impact depends on much more than just a technology innovation.

The most applicable HAS 2025 learnings reveal that while healthcare innovation continues to accelerate, its success depends on much more than technical breakthroughs. The future will be shaped by how well we navigate trust, equity, integration and mindset.

1. AI Is Only as Good as the Data Behind It

AI’s power comes not from algorithms alone, but from the data they depend on. Without clean, comprehensive and inclusive data sets, even the most sophisticated models fail to deliver consistent or equitable value. Healthcare’s long-standing issues with siloed data — from fragmented electronic medical records to unintegrated lab and claims data — remain a formidable obstacle.

Dmitry Tran, co-founder of Harrison.ai, emphasized that creating globally scalable healthcare AI requires including diverse patient populations in training, and guaranteeing imaging quality in every context. “Close enough isn’t good enough in healthcare,” he said. “Our AI must work equally well at UCLA and in a rural Vietnamese clinic.”

Our consensus: data access, standardization and representativeness are not back-end technical concerns; they are frontline healthcare challenges that directly shape outcomes.

Success in implementation depends not only on what’s possible, but on how organizational and behavioral forces align around what’s valuable and sustainable.

Howard Park, a data and AI advisor with deep experience in governance and deployment, reinforced this point by emphasizing that regulatory approval is only a starting point. Successful implementation, he noted, requires clarity around data diversity, technical integration and real-world user readiness.

2. Technology Doesn’t Fail, Adoption Does

Several panelists agreed that successful innovation isn’t about what’s built, it’s about what’s used. Katya Andresen, chief digital and analytics officer at the Cigna Group, noted that AI projects often stumble, not because of technical flaws, but because of weak adoption strategies. “Technology doesn’t fail. Change management does.”

AI implementation must start with clear communication and structured engagement. When clinicians and care teams are excluded from the design or rollout process, mistrust and friction follow. Technology leaders must focus not only on what AI can do, but also on how and why users should want it. “Organizations need to stop assuming buy-in,” Andresen added. “They have to earn it.”

3. Mindset Shifts Matter as Much as Model Shifts

AI adoption isn’t just about deploying new tools, it requires shifting how organizations think and operate. Khalil Smith, VP of inclusion and engagement at Akamai, reminded HAS attendees that human behavior is remarkably consistent over time: “Our brains haven’t changed in 10,000 years. We crave predictability. When change feels imposed, we resist it — even if it’s ultimately helpful,” he said.

Smith encouraged a behavioral lens for AI rollout: Involve users early, be transparent about intentions and impacts, and highlight how the technology helps not just the system, but also the individual. He also emphasized storytelling as a critical leadership tool. “People don’t remember data dumps. They remember narratives that connect with their experience.”

These shifts require leaders to go beyond managing adoption — they must model it.

4. Equity Must Be Designed into Innovation

AI does not automatically make healthcare more equitable. In fact, without intentional design, it risks amplifying the very disparities it promises to reduce. Tran underscored how higher-income patients — those likelier to be treated in academic medical centers and likelier to sign data-sharing consent forms — are disproportionately represented in training data sets.

The future will be shaped by how well we navigate trust, equity, integration and mindset.

The result? Tools that perform better for the already-privileged groups.

Panelists discussed solutions, including community partnerships, regulatory guardrails and research funding tied to diversity benchmarks. Smith noted that inclusive teams — diverse not only in demographics, but in training and perspective — are likelier to catch potential blind spots before they’re baked into product decisions. Equity, they agreed, cannot be retrofitted. It must be built in from the beginning.

5. Implementation Is the Hardest — and Most Overlooked — Step

Many breakthrough ideas in healthcare analytics fail not because they’re wrong but because they’re incomplete. “A great model without implementation is just a paper,” said Eran Halperin, chief AI officer at Miller Health and a professor of computer science at UCLA.

Halperin encouraged researchers and developers to think backward: Start from a clinical challenge, identify the organizational and technical steps needed for change, and then build the model. Implementation isn’t just the final mile — it’s most of the marathon.

Jake Schultz, managing director at Seniors Health and a former IBM Watson Health executive, echoed this perspective. He recalled the difficulty of scaling even promising AI pilots in the past: “Back then, we had great ideas, but terrible infrastructure and unclear incentives. Everyone wanted to test. No one wanted to own.”

But Schultz was also optimistic about where the field is heading. “Today, we’re seeing more traction because organizations have started to align around data, workflows and a clearer sense of what AI can realistically do. The groundwork is finally catching up to the vision,” he said. He emphasized the importance of connecting siloed systems in real time: “AI is only useful if it can pull from across fragmented systems to provide insight at the point of care, not after the fact.”

AI does not automatically make healthcare more equitable. Without intentional design, it risks amplifying the very disparities it promises to reduce.

Closing the implementation gap requires stronger operational partnerships, clearer incentives and leadership willing to sustain innovation beyond the demo phase.

Daniel Benjamin, professor of behavioral economics at UCLA Anderson, observed that misaligned incentives — rather than technical shortcomings — often stand in the way of scalable solutions. His perspective reinforced the idea that success in implementation depends not only on what’s possible, but on how organizational and behavioral forces align around what’s valuable and sustainable.

6. The Goal Is Human-AI Collaboration, Not Human Replacement

No HAS speaker advocated replacing healthcare workers with AI. On the contrary, many shared how AI, despite popular fears, is best used when it enhances human capacity.

Rob Alger, SVP at Kaiser Permanente, described how AI is transforming call center operations by automating routine lookups across dozens of outdated systems so representatives can focus on problem-solving and empathy. In training, AI-generated synthetic calls allow tailored learning experiences, enabling faster and more confident onboarding.

“AI should not be a wall between patients and providers,” Alger said. “It should be a ladder.”

This perspective echoed throughout the day: that the most promising use of AI in healthcare is not automation, it’s augmentation.

7. Trust Can Flip the Adoption Curve

Perhaps the most surprising insight of the symposium came from a real-world example: Tran shared that in one hospital, radiologists refused to read chest X-rays unless Harrison.ai’s tool was turned on. The AI had become a valued copilot, trusted not just to speed up diagnosis but to catch what humans might miss.

This example was about earned trust, built through transparency, strong performance and thoughtful integration. “That’s when you know adoption is real,” Tran said.

Andres Terech (Ph.D. ’04) is faculty director of UCLA Anderson’s Morrison Center for Marketing and Data Analytics and an adjunct professor of marketing. Recordings of HAS 2025 sessions are available on the Morrison Center website. UCLA Anderson’s next Healthcare Analytics Symposium will take place in May 2026.