AI is already in real clinical workflows—not just research papers.
The biggest near-term impact is earlier detection, faster drug pipelines, smarter use of existing medicines, and more precise treatment decisions.
No fully AI-designed drug has FDA approval yet as of mid-2026, but several have reached human trials with published results.
Drug discovery and design
- Insilico Medicine’s rentosertib is the clearest milestone: both the target and the molecule were generated by AI. It reached Phase IIa in idiopathic pulmonary fibrosis in about 30 months (typical timeline is 6–8 years).
- Patients on 60 mg daily gained 98.4 mL in lung function (FVC) vs. a 20.3 mL drop on placebo.
- Isomorphic Labs (DeepMind) has AI-designed candidates approaching human trials; DSP-0038, an AI-discovered Alzheimer’s compound, entered Phase I.
- Pfizer used modeling and simulation to narrow candidates for Paxlovid. Moderna and Merck used AI on each patient’s tumor to build a personalized melanoma vaccine tested in 1,100 people.
Using drugs we already have (repurposing)
AI knowledge graphs match existing approved drugs to rare or “untreatable” diseases.
- Dr. David Fajgenbaum’s Every Cure platform scored combinations across thousands of drugs and diseases. One patient with a life-threatening inflammatory disorder improved enough on an unconventional chemo/immunotherapy/steroid mix to reach a stem-cell transplant. Similar matching has pointed to unexpected uses (e.g., certain cancer or rheumatoid-arthritis drugs for other conditions).
- Earlier, BenevolentAI’s graph flagged baricitinib (a rheumatoid-arthritis drug) for COVID-19.
Cancer screening and treatment decisions
- The MASAI randomized trial (105,934 women): AI-supported mammography found 29% more cancers and cut radiologist workload ~44%. Transpara as a second reader raised sensitivity 8.4% and caught more interval cancers.
- DeepHealth’s FDA-cleared breast ultrasound tool improved cancer detection sensitivity ~8% and cut interpretation time 37%.
- Sutter Health’s Epic-integrated AI tracker roughly doubled early-stage lung cancer diagnoses; more than 70% of cases are now found at stage 1 or 2.
- Perimeter Medical’s Claire (FDA PMA) uses AI optical coherence tomography during lumpectomy so surgeons can check margins in real time and reduce repeat operations.
- Lab work has used generative AI (RFdiffusion and related tools) to design proteins that act as a “GPS” on T cells, helping them lock onto and kill melanoma cells.
Heart, stroke, and other acute care
- An ECG model trained on 10.6 million traces flags heart failure (~81%) and valve disease (~90%) in under two seconds; a multi-hospital trial is running toward possible NHS use.
- Aidoc, RapidAI, and similar systems triage CT for hemorrhage, large-vessel stroke, PE, and incidental findings so treatment starts sooner. RapidAI received multiple new FDA clearances for stroke and brain-injury modules.
- Google’s diabetic-retinopathy models have been used in more than a million screenings, with a diagnosis in as little as two minutes.
Chronic disease and mental health (already in Medicare pilots)
- Limbic’s Unpacked: clinician-supervised CBT delivered by an AI voice agent; first mental-health company in the FDA’s TEMPO real-world evidence pilot.
- Cadence HypertensionOS: AI-assisted, clinician-supervised med titration for Stage 2 hypertension (also in TEMPO).
- Emory used agentic AI phone follow-up to move blood-pressure control from 1-star to 4-star CMS rating.
Scale of regulation
The FDA authorized 258 AI medical devices in 2025 alone and more than 1,450 through 2025.
Radiology accounts for the large majority. Most clearances use existing evidence pathways rather than new randomized trials.
Bottom line: AI is already changing who gets treated, how fast, and with which drug or procedure.
The strongest current examples are earlier cancer detection, faster candidate drugs, and matching old drugs to new diseases.
Full “AI-cured-this-disease” headlines are still rare because approval and outcomes take years—but the pipeline and the clinic tools are no longer theoretical.
Taking Time is important unlike we did with Covid 19 forced shots.
