AI in Healthcare: What’s Actually Happening Right Now
Your work in healthcare will change more in the next 3 years than in the last 20. Here’s why.
Part of my job as a Biomedical Data Scientist is staying up to date with what’s happening in AI applied to medicine — week by week, it’s not optional. I decided that instead of keeping these notes to myself, it makes more sense to share them. So that’s exactly what this is: my weekly notes, organized to be useful to you, a few times a month.
What’s happening with AI in medicine right now is not a trend — it’s a genuine inflection point.
Why this matters now
Between July and August 2026, GPT-5.6, Grok 4.5, and Muse Spark 1.1 launched almost simultaneously, triggering a price war that cut inference costs by up to 10x [1][2][4]. DeepSeek V4 went open source. AMD and Astera Labs reported record Q2 2026 results with specialized hardware eliminating memory bottlenecks [8][9].
What six months ago was a large-hospital project is now viable for a mid-size clinic. The barrier is no longer economic. The barrier now is the willingness to act.
“Organizations that start now — even with a small pilot — will be years ahead of those still waiting for the perfect moment.”
The core idea: 5 areas where AI is working today
1. AI medical scribes
A multicenter study documented significant reductions in administrative time after deploying AI scribes [12][13]. Clinicians reclaim hours per day. ROI is measurable in weeks, not years.
2. Explainable imaging diagnostics
Radiomics combined with deep learning is transforming the evaluation of adrenal masses [11]. What’s new: the system explains its reasoning. Physicians don’t trust black boxes — this solves that barrier.
3. Personalized cardiovascular risk prediction
Current models outperform conventional ASCVD scores because they capture complex clinical interactions that traditional methods miss [14][17]. Available for immediate deployment.
4. AI agents for Alzheimer’s and dementia
Not chatbots. Agents that execute complete workflows — triage, coordination, follow-up — integrated with existing EHR systems in weeks [15]. The scalable solution for the neurological care capacity crisis.
5. Front-line health in Latin America
A public health agent designed for community health workers, accessible via mobile, without complex infrastructure [26]. AI for reproductive and maternal health in LATAM shows real promise but important maturity gaps [27][28].
The U.S. regulatory moment
HHS published a 2025 Request for Information on accelerating AI adoption in clinical care [21]. The FDA has three decades of AI/ML device authorizations — concentrated in radiology and cardiology [22][25].
The regulatory framework is being built right now. Organizations that engage will shape the rules. Those that wait will comply with someone else’s rules.
Oncology: the most active front
- Cancer Research UK committed £6 million for a translational research hub in Manchester [30]
- Africa is emerging as a player in precision oncology and functional genomics [31]
- The NIH INCLUDE project offers a replicable model: AI + common data elements for multicenter studies [33]
- OMOPCAN advances global standardization of oncological data using OMOP [34]
- The convergence of AI with multi-omics analysis opens new frontiers in cancer research [35][36]
What to do next
- Identify a repetitive administrative process in your organization — that’s where the easiest pilot starts.
- Define a measurable outcome before you start — time reduction, clinician satisfaction, error rate.
- Iterate from evidence, not intuition — AI models improve with structured feedback.
Final thought
What strikes me after 25 years in biomedical research is not the hype. It’s that peer-reviewed literature is now validating what pilots showed two years ago. The evidence cycle is shortening — that’s a maturity signal.
The perfect moment to start has passed. The second best moment is now.
If this was useful, follow for more notes on AI applied to public health, epidemiology, and biomedical data science.
References
[1] AI Breakthroughs 2026 — RealifeAI
[2] AI Breakthroughs July 2026 — KersAI
[4] August 2026 AI Mega Update — AIApps
[7] AI World Models & Continual Learning — NextBigFuture
[8] Semiconductors & AI Chips Aug 2026 — DistillIntelligence
[9] AI Weekly Newsletter Aug 2026
[11] Explainable AI for adrenal masses — PubMed
[12] AI medical scribes — multicenter study — PubMed
[13] AI-assisted clinical documentation — DOI
[14] AI in cardiology — cardiovascular risk — PubMed
[15] Agentic AI for Alzheimer’s and dementia — PubMed
[17] AI in echocardiography — PubMed
[21] HHS RFI — AI in clinical care — PubMed
[22] FDA three decades of AI/ML devices — PubMed
[25] AI in health — clinical applications — JAMA
[26] AI agent for front-line health workers — ResearchSquare
[27] AI reproductive/maternal health LATAM 1 — PubMed
[28] AI reproductive/maternal health LATAM 2 — PubMed
[30] Cancer Research UK £6M Manchester — Nature
[31] Africa — functional genomics in cancer — PubMed
[33] NIH INCLUDE — AI + CDE multicenter — PubMed
[34] OMOPCAN — OMOP in population oncology — PubMed
[35] AI + multi-omics in precision oncology — PubMed
[36] AI + multi-omics — book chapter Wiley
[38] Neurosymbolic robotics — arXiv April 2026