The Goal Isn't AI Usage. It's AI Impact: Lessons from ISPOR 2026

ISPOR 2026 had the unmistakable energy of a field picking up speed, and the conference felt like proof of it: a collision of ideas, sharp minds converging on the same questions, and real momentum building around what comes next. Conversations constantly circled back to AI, evolving policy, and a shared urgency around making HEOR smarter and more efficient. Walking away, I found myself with new perspectives on methodology, data-driven policymaking, and what it actually means to use these tools well. AI wasn’t just a topic of conversation at ISPOR this year, but the backdrop against which every conversation happened.  

ISPOR made it clear that the question is no longer if we should use AI, but how we can use it thoughtfully and efficiently. One phrase that cut through the noise and anchored everything was, “The goal is not AI usage, it’s AI impact.” Tied to that was a recurring theme: AI literacy is a prerequisite for AI use. Teams need fluency; they need to “speak AI” before it can meaningfully integrate into a workflow. But fluency alone is not enough. There was also a clear call to move past the idea of “human-in-the-loop,” a standard that no longer meets the moment. In HEOR and market access, humans need to be at the helm, because accountability will never fall on machines. It stays with us. That sense of ownership extends to how we think about progress in the field, and one such framework that stuck out to me was the AI maturity curve: moving from assistive, to integrated, to adaptive, to explainable. It’s a useful lens for understanding where the field is and where it is heading.

Right now, AI shines in tasks like literature reviews, abstract screening, large-scale data extraction, and workflow automation, areas where speed and scale matter and inputs are largely text-based. That last point is important: as large language models, they’re more reliable with words than with numbers. Numerical precision and statistical computation are areas where AI struggles, and deterministic tools like Excel remain better suited for those tasks. This is not a slight against AI, but a call to be realistic about what it’s built for. Understanding AI’s limitations and being optimistic about its potential are not in conflict; both are essential for meaningful AI innovation and widespread adoption. Part of that realism also means addressing a concern that came up more than once: a fear that AI will replace jobs. The most useful reframe from the conference speaks directly to that. AI won’t replace jobs, but people who use AI optimally will. AI won’t make things 100x faster, but it can help us answer 100x more questions. It is not about replacing what already works, but about expanding what is possible.  

While AI dominated the agenda, ISPOR covered a wide range of topics, and I made it a point to step outside AI-focused sessions and take in some of them. A few stood out. Social media listening generated real discussion: patient narratives are rich and authentic, offering genuine insight into lived experience, treatment efficacy, and accessibility. But scale doesn’t equal credibility. A lot of what circulates on social media is noise rather than signal, and the methodological and ethical questions around patient consent in this space are far from settled. Drug pricing was another area of focus, with an important reminder that price, spending, and value are not the same thing, and conflating them muddles the conversation. That same principle, that clarity of language shapes the quality of the conversation, carried into a discussion on communication in HEOR that deeply resonated with me. Evidence only matters if it can be understood. The human ability to translate complexity for different audiences isn’t going anywhere. If anything, AI can help us sharpen that skill, pushing us to become clearer thinkers and communicators rather than doing the thinking for us.    

The field is at a real inflection point. The organizations leading right now aren’t necessarily the ones moving the fastest, but the ones moving the most intentionally. There are meaningful gaps left to fill, including the need for standardized validation frameworks for AI applications specific to HEOR and market access. There is a growing consensus that AI should be treated not as a shortcut, but as a collaborator that requires clear workflows, thoughtful governance, and ongoing human judgement and oversight. ISPOR this year didn’t feel like a glimpse of the future. It felt like the starting line. The work of figuring out how to utilize AI well is just the beginning.