- Health AI Roundup
- Posts
- I read all the 2025 health AI predictions so you don’t have to
I read all the 2025 health AI predictions so you don’t have to
Arfa Rehman
January 06, 2025 • Reading Time: 4 minutes
It’s the first week of January and all our feeds are filled with 2025 predictions. I’ve read all the health-AI-specific ones, and there’s quite a few lofty takes like “this is THE year we’re going to see transformative widespread adoption of AI in healthcare!!!” To which I say…BFFR. This is healthcare we’re talking about!
So, in today’s special edition of this newsletter, I’m cutting through the noise to bring you 5 of what I think are the realest takes on what’s coming in 2025.
1. Focus will stay on low-risk use cases
AI scribing (automated documentation) and prior authorization are relatively low-risk areas for AI adoption because they focus on administrative tasks rather than high-stakes clinical decision-making. These tools address critical issues like clinician burnout and administrative bottlenecks, making them safer bets compared to clinical AI, which carries higher regulatory and patient safety risks.
“Ambient scribing and prior authorization are two areas where there may be a lot of traction for AI.”
AI will inch into clinical decision support with more proof-of-concepts, but the real action will stay in low-risk applications.
2. "Pilot Fatigue" will lead health systems to demand proof
Health systems are starting to get tired of AI pilots that don’t give a clear picture of ROI. From TechTarget’s predictions:
“Most health systems don't want to be the first to do this kind of work. They really want examples of organizations that have made it to value, that have demonstrated value, that later on, they can point to. They're looking for somebody that's established the path"
For startups, this means proven results will be necessary to gain traction in 2025.
3. Regulatory shifts are likely in 2025
Upcoming U.S. administrative changes could mean changes to AI regulations, according to the Fierce Biotech forecast:
“This year may bring major changes in how artificial intelligence shapes the healthcare industry—and vice versa—and tempering hype for the technology with ensuring patient safety will be essential.”
Until the uncertainty clears, we’ll likely see more focus on admin-focused AI, in line with prediction #1.
4. Divergence in AI models: Specialist vs. generalist
So far, we’ve primarily seen health AI applications relying on existing general-purpose models, which often run into limitations. This year, we may start to see a shift towards healthcare-specific models tailored to the industry’s unique demands, according to Out of Pocket predictions:
“There will be a divergence between generalized AI models and healthcare specific models as they sell into large healthcare enterprises. Healthcare models will focus much more heavily on accuracy, explainability, and security. ”
5. Most exciting AI developments may happen outside the clinic
Throwing my own prediction into the mix: while there’s a lot of focus on care delivery and clinician/patient-facing tools, the most exciting AI applications in 2025 could come from other, expanded applications in health and life sciences.
For example, Bluenote’s recent raise, as reported in our earlier newsletter this week, shows promise in the potential of AI being used to streamline regulatory processes for life science companies. We’re also seeing advancements in drug discovery and biotech manufacturing. This is an exciting but under-hyped area in my opinion, and may deliver the best AI developments this year.
That’s a wrap on the realest takes I’ve seen about 2025 predictions. While I see this year as a slow burn for health AI, it’ll nonetheless be an exciting one so if you want a front-row seat to it all, make sure to subscribe to Health AI Roundup!
Reply