Getting started: how to use AI to investigate your health

Most chronic health problems — gut issues, migraines, fatigue, persistent pain — are pattern-matching problems. Something triggers your symptoms, and your job is to figure out what. The challenge is that the relevant data is scattered across your meals, your sleep, your stress levels, and your daily life. No single doctor's visit captures it.

AI is useful here for a specific reason: it can hold all of that data in context simultaneously and look for correlations you'd miss manually. A primary care visit gives your doctor 15 minutes and whatever you remember to say. AI can compare six weeks of tracked data across a dozen variables and find the combinations that consistently predict your flares.

That's not a replacement for medical care. It's a different tool for a different part of the problem.

What AI does well

Cross-referencing data streams. Research on food-symptom tracking, published in PubMed Central, found that precise timing data reveals correlations that subjective recall consistently misses. AI extends this principle to every domain you track — sleep, stress, food, activity, mood — simultaneously.

Maintaining persistent memory. When you see a new provider, you re-explain your entire history. Iris stores your health context as organized notes — conditions, medications, test results, investigation findings — that carry across every conversation. Your investigation builds on itself rather than restarting. If Iris doesn't seem to have a specific piece of context, you can direct it: "load my notes on sleep" or "check my entries from this month."

Working at your pace. Doctors triage by severity. AI can dig into the specific details that matter to you — meal timing, sleep temperature, how you feel before your period — the things that get glossed over in office visits but might be the actual signal.

Cross-checking itself. Every message Iris sends gets reviewed by a separate Supervisor agent that checks for safety, overconfidence, and factual accuracy. If something's flagged, you'll see a supervisor review annotation appear a few seconds after the response. You can also visit the Supervisor tab to see conversation recaps and strategic reviews.

What AI doesn't do

Diagnose. AI can identify patterns in your tracked data. It cannot tell you what condition you have — that requires clinical evaluation, physical examination, and laboratory testing. Your goal is to use AI to gather evidence and walk into your provider's office informed.

Replace the hard work. Investigating your health takes time. A low-FODMAP elimination diet is 2-6 weeks. Sleep hygiene changes need consistent effort. AI can help you design and track these experiments, but you still have to run them.

Work without data. AI needs structured input — what you ate, when symptoms happened, how stressed you were. Iris handles the structuring for you (say "had a headache after lunch, 6/10" and it creates a structured entry with severity, timing, and triggers), but you still need to log consistently. Two weeks of daily tracking gives the Data Analyst enough entries to find real patterns. Vague or sporadic logging doesn't.

You need a plan

Without one, the common failure modes are: jumping from investigation to investigation without going deep enough, tracking data you never analyze, getting overwhelmed and quitting, or spending months investigating something a simple blood test could have ruled out.

The right plan depends on what you're dealing with, what you've already tried, and your capacity for tracking. Some people want detailed daily logs. Some people want to talk through their experience and have AI organize it. Both approaches work.

A good investigation plan is specific, realistic, and sequenced. It says "start here" — not "investigate everything."

References

1. Measuring Diet Intake and Gastrointestinal Symptoms in IBS — PubMed Central, 2020. Precise timing data revealing correlations missed by recall.
2. Patient engagement and health information technology — Health Affairs, 2016. Patient-generated health data improving clinical outcomes.
3. Biopsychosocial Model of Irritable Bowel Syndrome — PubMed Central, 2011. Multi-factor approach to chronic condition investigation.