Before HLTH Talks: Voices of Change – Fawad Butt
Building the Right AI Foundation
The conversation around AI in healthcare often focuses on shiny new tools and models. However, Fawad Butt, CEO of Penguin AI and former Chief Data Officer at Kaiser, UnitedHealthcare, and Optum, believes the real story lies in the foundation: governance, infrastructure, and the complex reality of data itself.
As we head into HLTH 2025, Fawad shared what leaders need to hear about AI adoption.
Generative AI Needs Modern Infrastructure
AI can’t be bolted onto outdated systems. Fawad predicts a wave of modernization as health systems and payers upgrade their core infrastructure to support cloud computing, generative AI, and automation. Without these upgrades, organizations will be stuck with biased, incomplete, or insecure data.
Real-Time Data Exchange is Here
Batch updates are fading. Thanks to FHIR and the 21st Century Cures Act, healthcare is moving toward real-time data exchange. This will be especially impactful in areas like prior authorization, where the new CMS regulations (effective January 2027) will mandate API-based exchange between patients, providers, and payers.
Smaller, Smarter AI Models
The future of healthcare AI isn’t massive general-purpose LLMs. It’s specialized, task-based models trained on trusted, regulatory-grade data. Organizations like Mayo Clinic and UPMC are leading by giving innovators access to de-identified, diverse, longitudinal datasets. This shift will create models that actually work in healthcare.
Bad Data = Bad Decisions
AI isn’t a magic wand.
“Without the right foundation, AI won’t fix bad data; it will just make bad decisions faster.”
Healthcare’s dirty secret is that much of its data is incomplete, inconsistent, and low quality. Unless organizations invest in cleaning and governing their data, AI will only accelerate poor decision-making.
Overlooked Opportunities
Fawad highlighted two critical but underutilized areas:
Multimodal data: medical images, voice, and text, which add real depth but remain difficult to standardize.
Phenotypical data: lifestyle, behavioral, and environmental inputs from wearables and census data, which can make AI more predictive and help detect risks earlier.
Together, these streams could unlock a more complete picture of health.
Change Management is the Hardest Part
Despite the headlines, the biggest obstacle isn’t the technology—it’s the people.
“People aren’t afraid of new tools; they’re afraid of change.”
Adoption requires investment in training, communication, and cultural readiness. Without it, the most promising technologies will never stick.
Why This Matters
As HLTH 2025 approaches, Fawad’s perspective is a timely reminder: AI won’t transform healthcare unless we first address the foundations of data, infrastructure, and governance. Leaders who invest in those fundamentals will be the ones who see AI deliver real-world value.
🔥 This is the fourth installment of my HLTH Talks: Voices of Change series. Stay tuned for more candid conversations with the leaders shaping the future of healthcare.
👉 Subscribe now to follow the series and join the conversation.
#HLTHTalks #VoicesOfChange #HLTH2025 #HealthcareAI #HealthData