I Gave AI My Lab Results, My Prescriptions, and 30 Supplements. What It Told Me No Doctor Ever Had.
I am not a doctor. I want to say that at the start, not as a legal disclaimer, but as context for why what happened next surprised me.
I have spent 25 years making decisions inside large organizations — reading data, mapping dependencies, catching the thing nobody mentioned in the meeting that would break everything six weeks later. I thought I was reasonably good at holding complexity in my head.
Then I gave an AI my actual health data and asked it to audit my supplement stack. What came back made me rethink what I understood about the technology — and about where the real opportunity in AI is going to be found.
What I Actually Handed It
This wasn't a vague question. I gave it everything:
Labcorp results: HbA1c 5.8 (prediabetic range), testosterone 431 ng/dL, fasting glucose 95-103 mg/dL, LDL 102, triglycerides 120, HDL 43, MCH 26.3 (below range)
DEXA body composition: 39.1% body fat, 51.4% android fat distribution, 5.47 lbs visceral fat
Access Labs hormone panel: cortisol 9.1 ug/dL, TSH 1.51
Active prescription: Losartan 50mg for blood pressure
My existing supplement stack: approximately 20 items across capsules, powders, and teas — CoQ10, Fish Oil, Vitamin D3+K2, Black Seed Oil, Curcumin, Benfotiamine, Alpha-Lipoic Acid, Magnesium Glycinate, Magnesium L-Threonate, L-Theanine, NAC, Ginkgo/Brahmi/Shankapushpi, Boswellia, Joint Support, Gentle Iron, Triphala tablets, Calcium, Multivitamin
Eight new Ayurvedic supplements I was considering adding: Bacopa powder, Tulsi tea, Shatavari powder, Karela (Bitter Gourd) powder, Triphala powder, Ashwagandha powder, Jamun powder, Amla powder
I asked it to analyze each new item against my labs, my prescription, and my existing stack — and tell me what was safe, what to skip, and what to watch for.
What It Caught That I Would Have Missed
The hypoglycemia stack nobody warned me about
I was adding Berberine (an evidence-backed blood sugar reducer), Karela powder (Polypeptide-P that mimics insulin), Jamun powder (alpha-glucosidase inhibitor), Tulsi tea (glucose-lowering adaptogen), and Triphala. All natural. All individually researched. All individually reasonable for someone with HbA1c 5.8.
What the AI did was count them — and then count what I already had: Alpha-Lipoic Acid and Benfotiamine already in my stack also lower blood glucose. That made seven simultaneous glucose-lowering agents. With fasting glucose at 95-103 mg/dL, the distance to hypoglycemia threshold at 70 mg/dL is not large.
No single supplement website mentions this. A doctor seeing me for 15 minutes wouldn't know my full stack. The AI cross-referenced all seven simultaneously and told me the risk was real — especially dangerous, it noted, during CPAP sleep when I wouldn't wake from symptoms like dizziness or cold sweats.
It then gave me a mandatory sequenced introduction schedule: one new herb per week, with specific glucometer readings and stop-if thresholds at each stage.
The supplement that would have lowered my testosterone
I was considering Shatavari powder based on general Ayurvedic recommendations I'd read. The AI flagged it immediately: Asparagus racemosus is a female reproductive rasayana. Its phytoestrogenic saponins — specifically shatavarosides — can suppress LH and further reduce testosterone in men. My T at 431 ng/dL is already low-normal for my age. It would have moved in the wrong direction.
I would not have known this from the product page. I almost certainly would not have been told this in a standard medical visit. It saved me from a compound I was genuinely planning to buy.
The duplicate I was about to stack
Bacopa powder — which I was considering adding — is the same herb as Brahmi, which I was already taking in tablet form. Double-dosing Brahmi raises GI upset risk and can slow heart rate. The AI identified the botanical overlap (Bacopa monnieri = Brahmi) across different product names and flagged it as a duplicate before I spent money or created a problem.
The antiplatelet stack I was building without knowing it
My existing stack already included Fish Oil, Ginkgo, Black Seed Oil, and NAC — all with antiplatelet properties. The new additions (Amla, Tulsi, Karela, Triphala's Haritaki component) would add four more. The AI flagged that before any dental procedure, extraction, or surgery, I would need to stop all eight for 7-10 days and inform the provider. Not because the risk is catastrophic — I'm not on warfarin — but because I would never have told a surgeon about my fish oil and Amla powder without knowing they were anticoagulant in combination.
What AI Cannot Do
I want to be honest about the limits, because the limits matter.
The AI cannot examine me. It cannot order the ferritin test it correctly identified I needed before continuing iron supplementation. It cannot adjust Losartan. It cannot diagnose the cochlear issue it flagged as urgent. Every output it gave me came with a clear instruction: review with Dr. Lamb, your Ayurvedic practitioner, and your ENT. That instruction is real, not decorative. I have an appointment.
What it produced is not a prescription. It is the most rigorously prepared briefing document I have ever brought into a medical appointment.
Why This Matters as an Investment Thesis
I have spent 25 years watching technology get adopted inside large organizations. There is a pattern: the first people to dismiss a technology are often those whose expertise it displaces, and the first people to overestimate it are those who haven't seen where it fails.
What happened with my supplement protocol sits in neither category. This was AI operating in its genuine zone of competence: synthesizing large volumes of structured information, holding multiple variables simultaneously, cross-referencing across domains (Ayurveda, pharmacology, endocrinology, hematology), and identifying the interaction that emerges only when you look at the full system rather than any single piece.
A general practitioner sees patients for 15 minutes. A specialist knows one domain deeply. A pharmacist catches drug interactions but doesn't know your supplement stack unless you declare it, and most people don't. An integrative medicine physician can do more of this — but the appointment costs time and money and access that most people don't have.
What AI is doing in the health space is not replacing the physician relationship. It is making it possible for people who cannot afford concierge medicine — or who simply don't have a doctor who knows Ayurveda, pharmacology, and endocrinology fluently — to walk into a medical appointment with a document that shows their full picture.
That is not a small thing. Access to rigorous, personalized health information has historically been a function of wealth and geography. If you had a brilliant friend who happened to be a pharmacist, an integrative medicine doctor, and an Ayurvedic practitioner simultaneously, and who had the time to read all your labs before you asked a question — you would use that friend constantly. Most people don't have that friend. This is beginning to approximate that.
The investment opportunity I see is not in the AI model itself. The model is infrastructure, like the database or the cloud. The opportunity is in the applications that are building on top of it for domains — health, law, financial planning — where the information asymmetry between expert and patient has always been widest and most consequential.
The companies that figure out how to put genuinely rigorous, personalized, accountable AI into the hands of people who couldn't previously access expert-level guidance in these domains will build enormous businesses. The early signals are already visible if you know what you're looking at.
What I'm Actually Doing With This
My protocol is now v2.0. Six of the eight new Ayurvedic herbs are being added in a phased sequence starting this week — one at a time, with daily glucometer readings and weekly blood pressure checks. Two (Shatavari and Bacopa) are off the list. I've printed the protocol document for my next appointment with Dr. Lamb.
The most honest summary: I have never brought a better-prepared document into a medical appointment. And I have never had a clearer sense of what I'm putting into my body, why each item is there, and what to watch for.
That's not nothing. In fact, it's close to everything.
The full supplement protocol document is attached. This article is a personal account of using AI as a research and synthesis tool. It is not medical advice. Work with your physician before making changes to any supplement or medication regimen.
Satya Sivunigunta is a former CEO of SelectBlinds ($220M ecommerce), with prior operating roles at Nike, Microsoft, Ogilvy, JCPenney, and Conn's HomePlus. He writes about operating, leadership, and technology at satya.me.
What would you hand AI if you thought it could actually read it?