I Built a GLP-1 Telehealth Clinic and Lost Everything When the FDA Made It Unviable. Here's What the Medvi Story Doesn't Tell You.

By Prahlad Menon 7 min read

This week, The New York Times profiled Matthew Gallagher and his company Medvi — a GLP-1 telehealth startup he built from his house in Los Angeles with $20,000, more than a dozen AI tools, and one employee (his brother). In 2025, its first full year in business, Medvi generated $401 million in sales. In 2026, the company is on track to hit $1.8 billion. 250,000 customers. A 16.2% net margin. Two employees.

I launched the same business in November 2024.

Five months later it was over.

This is the post-mortem. Not the version where I pretend the timing was unforeseeable and the outcome was inevitable. The honest one.

The Model We Were Both Running

The core business is simple: connect patients who want GLP-1 weight loss medications (semaglutide, the active ingredient in Ozempic and Wegovy) with a telehealth physician who can prescribe them, then fulfill via a compounding pharmacy at a fraction of the brand-name price.

Why compounding? Because brand-name Ozempic runs $1,000+ per month out of pocket. Compounded semaglutide — chemically identical active ingredient, manufactured by a licensed 503A or 503B pharmacy — could be offered for $150-300/month. That price gap is what made mass-market patient acquisition possible. Without it, you’re selling to people with great insurance or significant disposable income. The market shrinks by an order of magnitude.

It’s worth noting that the insurance landscape has also shifted meaningfully since 2024. More insurers are now willing to cover GLP-1 medications — which sounds like good news for patients, but it actually changes the cash-pay telehealth model in a complicated way. Patients who have insurance coverage go through their insurer, not a cash-pay telehealth clinic. The cash-pay market now skews toward people whose insurance still won’t cover it, which is a narrower and harder-to-reach segment than the broad “can’t afford Ozempic” market that existed when compounding was the only affordable option.

Gallagher built Medvi on this model. I built my clinic on this model. So did hundreds of other telehealth operators during 2023-2024.

This is the part most coverage breezes past. Compounding pharmacies can’t normally produce an “essentially a copy” of an FDA-approved drug — that’s a violation of the Food, Drug, and Cosmetic Act. The exception: when a drug appears on the FDA’s drug shortage list, the commercial availability restriction doesn’t apply.

Ozempic and Wegovy faced genuine nationwide shortages starting in 2022 as demand for GLP-1 medications exploded following high-profile clinical results. The FDA placed semaglutide on the shortage list. That single regulatory classification opened the door for compounding pharmacies to legally supply the telehealth market — and the entire industry was built inside that window.

The Timeline That Actually Matters

Here’s the sequence, with exact dates, sourced from regulatory filings:

2022–2023: Semaglutide placed on FDA drug shortage list. Compounding pharmacies (503A patient-specific and 503B outsourcing facilities) legally begin large-scale production. GLP-1 telehealth clinics proliferate.

September 2024: Matthew Gallagher launches Medvi. 300 customers in month one. 1,000 more in month two.

November 2024: I launch my clinic. Roughly $3,000/month in fixed costs — staff, marketing, operations.

February 21, 2025: FDA updates the drug shortage list and declares all doses of injectable semaglutide (Ozempic/Wegovy) as no longer in shortage. The legal foundation for bulk compounding begins to collapse.

April 22, 2025: Enforcement deadline for 503A compounding pharmacies. As documented by Burr & Forman’s legal analysis: “503A compounding pharmacies may continue to compound semaglutide injection products until April 22, 2025.” After this date, compounded semaglutide that is “essentially a copy” of the FDA-approved drug becomes legally untenable.

May 22, 2025: Same deadline for 503B outsourcing facilities.

My clinic: Never reached traction. With the compounding window closing and acquisition unsolved, the economics collapsed. Gone.

September 9, 2025: FDA and HHS launch a broad enforcement initiative specifically targeting telehealth providers marketing compounded GLP-1s. Warning letters sent to operators still advertising claims like “same active ingredient as Ozempic” and “generic Ozempic.” The window isn’t just closing — it’s being enforced shut.

April 2026: NYT publishes the Medvi story. Gallagher’s revenue: $401M in 2025, on track for $1.8B in 2026.

How Medvi Actually Works: The Full Stack

This is the part the NYT piece glosses over and most coverage misses entirely. Medvi is not a healthcare company. It’s a distribution layer built on top of rented healthcare infrastructure. Understanding the stack explains both how he built it so fast and why it’s more fragile than the $1.8B number suggests.

Layer 1: The Customer Interface (Medvi itself)

Medvi owns the brand, the website, the patient onboarding flow, and the marketing. That’s it. Everything the patient sees and touches — the landing page, the intake questionnaire, the ad they clicked — is Medvi. AI tools generated most of it: ChatGPT, Claude, and Grok for code and copy; AI-generated ad creative including video; an AI chatbot for customer service. The entire AI tool stack reportedly costs $3,000–$12,000/year.

Layer 2: The Physician Network (OpenLoop Health)

Medvi has no doctors. OpenLoop Health is a “telehealth-as-a-service” platform that provides licensed physicians across all 50 states on demand. OpenLoop handles the clinical consultations, prescribing, state licensing compliance, and malpractice coverage. Gallagher is quoted in the NYT piece directly: “He could use A.I. to do the branding and marketing and let CareValidate and a similar platform, OpenLoop Health, handle the doctors, pharmacies, shipping and compliance.” The physicians are, as the Pupuweb analysis put it, treated as “compliance endpoints — essentially physical-world API calls that satisfy a legal checkbox.”

Layer 3: Prescription Routing (CareValidate)

CareValidate handles prescription verification and routing — ensuring prescriptions are valid, properly documented, and sent to the right pharmacy partner. This is the compliance glue between the physician layer and the pharmacy layer.

Layer 4: Compounding Pharmacies

Medvi has no pharmacy. Multiple unnamed 503A/503B compounding pharmacy partners manufacture and ship the medications directly to patients. Medvi’s own website states: “MEDVi does not produce compounded medications.” The pharmacies take on the manufacturing liability. Medvi takes the margin.

The AI Operations Layer (Workflow Agent Fit)

What Gallagher actually built internally — beyond the customer-facing layer — is what he calls “Workflow Agent Fit”: a system where lead generation, medical intake, and pharmacy routing are chained together through APIs and AI agents with minimal human intervention. Lead comes in → AI qualifies → intake form sent → physician consultation scheduled via OpenLoop → prescription issued → CareValidate routes to pharmacy → pharmacy ships. Each step is an API call, not a human handoff.

The automation also includes a three-system redundancy setup where independent AI agent clusters monitor each other’s outputs and “outvote” errors — his fix for the known problem that AI customer service bots hallucinated drug prices and generated fake “before and after” photos early on.

The Honest Assessment of the Stack

Forbes noted what Gallagher himself has acknowledged: “Medvi holds no proprietary technology, no licensed physician network, no pharmacy infrastructure, and no exclusive supplier relationships.” Every critical component is rented and replaceable — by Gallagher or by a competitor. The moat is the brand and the acquisition machine, not the infrastructure.

This is the model’s strength and its vulnerability simultaneously. It scales fast because there’s nothing to build. It can be replicated fast for the same reason.

What Gallagher Got Right That I Didn’t

1. He cracked acquisition before the window closed

The NYT piece describes Medvi’s first month: 300 customers. Second month: 1,000 more. That’s the entire story. Not the AI stack, not the vendor infrastructure — the fact that he solved paid patient acquisition for a healthcare product on platforms that actively restrict healthcare advertising.

That’s the hard part. Google, Meta, and Instagram all have layered restrictions on health and weight loss advertising on top of FDA rules. Getting a compliant, converting campaign in that environment is genuinely difficult. Gallagher apparently cracked it fast. I didn’t crack it at all in five months.

Without acquisition, every other piece of the model — the clinical infrastructure, the AI ops layer, the pricing — is irrelevant.

2. He built lean enough to survive on the margin

Two employees. AI handles customer service, content, code, ad creative, business analysis. The NYT framing is “AI built this company” — which is reductive, but the cost structure point is real. A business running on $3K+/month in fixed overhead needs to reach breakeven fast. A business running on near-zero marginal cost per additional operation has much more runway to find acquisition product-market fit.

I was running a heavier operation than the model required.

3. He launched early enough to have momentum before April 22

Gallagher launched September 2024. By April 22, 2025 — seven months later — he had 250,000 customers, established brand recognition, and the resources to pivot toward brand-name medication pathways as compounding restricted. The regulatory change was a headwind for him. For me, at month five with no traction, it was a kill shot.

What the Medvi Coverage Gets Wrong

The framing in most coverage — including the NYT piece — is that this is primarily a story about AI enabling a new kind of company. That’s a real and interesting trend, but it obscures the more important structural story.

Medvi succeeded in a specific regulatory window that is now largely closed. The compounded semaglutide market that enabled the $99-$150/month price point — the price point that drove 250,000 customers — no longer exists in the same form. Anyone reading the Medvi story as a template for “build a GLP-1 clinic with AI tools in 2026” is misreading the situation.

There’s a double squeeze happening now: compounding is restricted and insurance coverage has improved. The patients most likely to convert to cash-pay telehealth — those who want GLP-1s but can’t get them through insurance — are a shrinking and more price-sensitive group than the 2024 market. Gallagher is reportedly pivoting Medvi toward brand-name pathways and insurance-compatible models. That’s a very different business than the one the NYT piece describes.

As Holland & Knight noted in their September 2025 analysis of the FDA enforcement action: “The telehealth industry saw and participated in the emergence and drastic growth of the GLP-1 marketplace while Semaglutide and Tirzepatide were on the FDA’s drug shortage list… Semaglutide and Tirzepatide are now off the drug shortage list, and the standard 503A and 503B pharmacy rules apply.”

The $401M Medvi story is a window-timing story as much as an AI story.

What’s Still Possible in Telehealth

The structural insight behind GLP-1 telehealth remains valid: there is a massive underserved middle between expensive in-person specialty care and ineffective supplements, and direct-to-consumer telehealth can serve it profitably if you own the patient acquisition channel and own the patient acquisition channel.

The GLP-1 window is largely closed. But the same model applies to other categories with better regulatory stability:

Cardiovascular risk reduction. Statins, blood pressure management, preventive cardiology. The drugs are cheap, generic, off-patent — no compounding dependency, no shortage list risk. Clinical protocols are well-established. Heart disease remains the leading cause of death in the United States. Primary care is overwhelmed and most patients who should be on lipid-lowering therapy aren’t. My background — PhD from Carnegie Mellon, work at the intersection of AI and cardiology — gives me credibility here that a pure-distribution founder can’t replicate.

Metabolic health beyond GLP-1. Thyroid management, metabolic panels, insulin resistance — categories with similar demand but regulatory profiles that don’t depend on shortage exceptions.

The AI cost structure changes the math for 2026. Patient intake, onboarding, follow-up messaging, refill reminders, FAQ responses — all of this runs on AI now at near-zero marginal cost. The staffing cost structure that burned my clinic can be eliminated almost entirely. Combined with a leaner operational model, the breakeven timeline changes dramatically.

The Honest Conclusion

I got the market right. I got the model right. I launched at the right time in the right window. I didn’t crack acquisition fast enough, I ran a heavier cost structure than necessary, and I ran out of runway five months before the regulatory environment shifted anyway.

That’s a different failure than having the wrong idea. It’s also a more useful failure to understand — because the same regulatory clock that killed my clinic is what makes the Medvi story remarkable rather than just repeatable.

The question worth asking now isn’t “how do I build the next Medvi?” The question is: what clinical category has the same demand profile as GLP-1 weight loss, with a regulatory foundation that doesn’t depend on a temporary shortage exception?

That’s the question I should have started with in 2024. It’s the question I’m building around now.

This post was also discussed on r/EntrepreneurRideAlong.

Sources:

Prahlad G. Menon, PhD, PMP is the founder of QuantMD and ThinkCreate.AI, working at the intersection of AI and healthcare.