In partnership with
AI Agents Are Reading Your Docs. Are You Ready?
Last month, 48% of visitors to documentation sites across Mintlify were AI agents—not humans.
Claude Code, Cursor, and other coding agents are becoming the actual customers reading your docs. And they read everything.
This changes what good documentation means. Humans skim and forgive gaps. Agents methodically check every endpoint, read every guide, and compare you against alternatives with zero fatigue.
Your docs aren't just helping users anymore—they're your product's first interview with the machines deciding whether to recommend you.
That means:
→ Clear schema markup so agents can parse your content
→ Real benchmarks, not marketing fluff
→ Open endpoints agents can actually test
→ Honest comparisons that emphasize strengths without hype
In the agentic world, documentation becomes 10x more important. Companies that make their products machine-understandable will win distribution through AI.
Hi!
After a few weeks in Mexico, I’m back in the U.S. Over the past few weeks, everyone in my family got sick at least once (including me 🤒). I’m hoping the rest of the season is a much healthier one for all of us.
It also feels good to be back at the Neurobionics Lab. Lately, we’ve been working on some exciting robotic projects. I’ll share more when the time is right.
Now, let’s dive into today’s issue.
🤖 AIBytes
Researchers tested a medical AI called AMIE (Articulate Medical Intelligence Explorer) in a real clinic. The goal was to see if the AI could ask patients questions, suggest possible diagnoses, and help doctors prepare for visits.
🔬 Methods
Ambulatory urgent care at an academic medical center.
Participants 100 adult patients (98 completed).
Patients chatted with AMIE before their appointment
The AI asked about:
symptoms
medical history
The system created:
a list of possible diagnoses
a suggested care plan
Safety: Human supervisors watched every conversation.
Evaluation:
Doctors later reviewed the AI results.
Researchers checked the final diagnosis from medical records after 8 weeks.
📊 Results
Safety: No conversations had to be stopped.
Diagnostic accuracy
The correct diagnosis appeared in the AI list in 90% of cases.
The correct diagnosis was in the top three suggestions in 75% of cases.
Comparison with physicians
Experts reviewing the results found no meaningful difference between AI and physicians for:
diagnostic reasoning
care plan safety
care plan appropriateness
Physicians did plans significantly better:
more practical
more cost-effective
Patient feedback:
About 90% completed the survey
75% said the AI helped them prepare for the visit
🔑 Key Takeaways
A large language model can safely collect medical history from real patients.
The AI suggested the correct diagnosis in most cases (90%).
Diagnostic reasoning was similar to physicians in blinded review.
However, physicians still created more practical care plans.
🔗 Brodeur P, Koshy JM, Palepu A, et al. A prospective clinical feasibility study of a conversational diagnostic AI in an ambulatory primary care clinic. https://arxiv.org/pdf/2603.08448
Researchers developed HealthGuide@Home, an AI system that generates personalized diet and exercise plans for residents enrolled in Singapore’s Healthier SG preventive care program.
🔬 Methods
Participants: 20 residents (age 40–59) and 7 clinicians.
AI system: Agent-based platform using a LLM with a semantic router.
The system used patient demographics, health conditions, preferences, and contextual information to generate personalized diet and exercise plans.
Residents evaluated the plans using a 1–5 scale for:
appropriateness
usefulness
actionability
personalization
Acceptance threshold: median ≥3.0.
📊 Results
Model performance: The new system scored slightly higher in reasoning, function calling, and instruction following.
Answer quality:
The first model was more accurate.
The second model was slightly more relevant.
Resident ratings: Scores were high across all domains (median ~3.8–4.0 / 5).
Most residents preferred personalized plans and detailed guidance.
No safety concerns were identified.
🔑 Key Takeaways
Agent-based AI systems can generate personalized preventive health plans.
Users rated AI recommendations useful, appropriate, and actionable.
Personalization and detailed guidance were highly valued by participants.
Larger studies are needed to determine clinical outcomes and long-term behavior change.
🔗Goh, H.L., Sancenon, V., Chu, B.M.X. et al. Personalised health plan development using agentic AI in Singapore’s national preventive care programme: a pilot study. npj Digit. Med. (2026). https://doi.org/10.1038/s41746-026-02514-8
🦾TechTools
Handheld ultrasound device that connects to a smartphone for point-of-care imaging.
Uses AI guidance to help clinicians capture diagnostic-quality images.
Portable and used across specialties including emergency medicine, cardiology, and primary care.
AI ambient documentation platform designed for emergency and acute care clinicians.
Listens to doctor-patient conversations and automatically generates structured clinical notes.
Provides real-time insights, risk alerts, and billing documentation to improve workflow and reduce charting time.
AI video platform that creates realistic avatar videos from simple text.
Supports 140+ languages and voices, no camera or recording needed.
Useful for patient education, training, and medical presentations.
🧬AIMedily Snaps
Policy Issues for Integrating AI in Cancer Research and Care: A Workshop (Link).
Stanford: AI platform maps disease risk from space (Link).
7 clinical studies to validate AI technologies for minimally invasive brain and cancer treatments (Link).
J. Hopkins: AI-Based liquid biopsy may detect liver fibrosis, cirrhosis and chronic disease dignals (Link).
Amazon Connect Health to reduce administrative burden in health care (Link).
Automated CT scan analysis could fast-track clinical assessments (Link). Original paper ⬇️.
🧪Research Signals
Decoding the language of sleep with AI (Link).
Vision-Language System using Open-Source LLMs for consent and instruction gestures in medical interpreter robots (Link).
Merlin: a CT vision–language foundation model and dataset (Link).
Model Medicine: A clinical framework for understanding, diagnosing, and treating AI models (Link).
A Framework for healthcare from the eye: Oculomics as a powerful window to systemic health (Link).
🧩TriviaRX
A surprising clue helped researchers discover potential biomarkers for Parkinson’s disease. What unusual signal first alerted scientists to investigate this idea?
A) Changes in skin odor noticed before diagnosis
B) Subtle tremor patterns detected by smartphones
C) Variations in pupil size during eye exams
D) Changes in walking rhythm measured by wearables
Now, let’s see the correct answer from our last TriviaRX:
IDx-DR is an autonomous AI system approved in 2018 that detects diabetic retinopathy from retinal images and recommends referral without requiring clinician interpretation.
We’re done for today.
As always, thank you for taking the time to read.
You’re already ahead of the curve in medical AI — don’t keep it to yourself!
Forward AIMedily to your colleagues who’d appreciate the insights.
Until next week.
Itzel Fer, MD PM&R
Forwarded this email? Sign up here
P.S. Enjoying AIMedily? ✅ Write a quick review here
How did you like today's newsletter?
1
