Basics
| Item | Details |
|---|---|
| Founders | Matthew Gallagher (CEO) + Elliot Gallagher (his brother) |
| Product | Medvi — GLP-1 telehealth platform (weight-loss prescriptions) |
| Team size | 2 people |
| Revenue | $401M first year (2025); 2026 target $1.8B |
| Net margin | 16.2% vs. Hims & Hers (2,442 employees) 5.5% |
| Customers | 250,000 (first year) |
| Starting capital | $20,000 |
| Zero to live | 2 months (launched September 2024) |
| AI stack | ChatGPT + Claude + Grok (code / copy), Midjourney (ad creative), Runway (video ads), ElevenLabs (voice support), in-house AI Agent |
| Source | PYMNTS / Inc / MirrorReview |
Storyline
Background: Matthew Gallagher saw the explosive demand for GLP-1 weight-loss drugs (Ozempic, Wegovy, etc.) — over 40% of U.S. adults are obese, and the traditional clinical route is slow and expensive. His core insight wasn’t “build an AI product” but “use AI to replace an entire team and run an extremely lean telehealth platform.”
Turning point: Matthew adopted a radical “rented infrastructure” strategy — outsource all regulated medical functions (physician supervision, prescription issuance, drug fulfillment) to compliance platforms like CareValidate and OpenLoop Health, and keep only the customer-facing layer in-house: brand, website, checkout, marketing. On top of that, he used the AI stack to replace what would traditionally need a 20–50 person team: ChatGPT/Claude/Grok for code and copy, Midjourney for ad creative, ElevenLabs for voice support, an in-house AI Agent for daily ops. Total starting capital: $20,000. Live in 2 months.
Outcome: First-year (2025) revenue $401M, 250,000 customers, 16.2% net margin. For comparison, peer Hims & Hers has 2,442 employees and $2.4B in revenue but only a 5.5% margin. Medvi reached close to 1/6 of that revenue with 2 people — and 3x the margin. The 2026 target is $1.8B — if reached, it will be the real-world case study for Sam Altman and Dario Amodei’s “one-person billion-dollar company” prediction.
Workflow Breakdown
Minimum loop analysis
| Step | What he did | AI involvement | Notes |
|---|---|---|---|
| 1. Think it through | Identified the GLP-1 demand surge + the chance to outsource telehealth compliance | Low | Business insight is fully human |
| 2. Write a spec | Minimal product definition: signup → questionnaire → physician remote review → prescription → delivery | Medium | AI assists in laying out user flows and compliance requirements |
| 3. Implement | ChatGPT/Claude/Grok generate site code + AI Agent automates ops | High (~85%) | Code, copy, ad creative are almost fully AI-generated |
| 4. Validate | Real paying users — 250K in year one | Medium | User behavior analysis may use AI assist |
| 5. Iterate | Continuously optimize acquisition channels and AI Agent efficiency | High | Midjourney generates ad variants at scale; AI optimizes spend |
Tool stack
- ChatGPT + Claude + Grok: three tools used in parallel, covering code and copy — no single-AI dependency, pick the best tool per task.
- Midjourney (ad creative): mass-generates ad image variants, replacing a 2–3 person design team. Dozens of variants tested per day.
- Runway (video ad generation): AI video ads replace a 3–5 person video team.
- ElevenLabs (voice support): AI voice support replaces a traditional call center, 24/7 automated.
- CareValidate + OpenLoop Health (outsourced infrastructure): not AI tools, but the core of the business model — outsource the regulated heavy assets, keep the asset-light layer.
Roles one person carried
| Role | AI replacement | Notes |
|---|---|---|
| Development / tech | 85% | ChatGPT/Claude/Grok generate nearly all code |
| Design / advertising | 90% | Midjourney + Runway replace an entire creative team |
| Customer support | 90% | ElevenLabs voice + AI Agent handle most of it |
| Marketing | 70% | AI generates ad creative, but media-buying strategy and budgets are human |
| Compliance / medical | Outsourced | CareValidate + OpenLoop Health handle all regulated functions |
| Strategy / decisions | 0% | Business direction and expansion pace are fully Matthew’s |
3 Reproducible Key Moves
-
“Rent infrastructure” instead of building it: Medvi’s core innovation isn’t tech — it’s business architecture. Outsource heavy assets (medical compliance, prescriptions, fulfillment) entirely; keep only the asset-light customer-facing layer. The model applies to any regulated industry: fintech (use Stripe Treasury to outsource banking licenses), education (use Coursera/Udemy to outsource course infrastructure). How to do it: map your business flow, tag each step’s compliance requirements and cost. Outsource everything you can, keep only what directly creates user value.
-
Run AI tools in parallel, don’t bet on one: Matthew uses ChatGPT, Claude, and Grok at the same time — not because he can’t tell which is best, but because different tasks have different optimal answers. How to do it: Claude is good at long content / code review, ChatGPT is good at creative ideation, Grok is good at real-time info. Pick the tool by task — don’t be “religious” about one.
-
Replace brand-building with ad automation: Medvi didn’t Build in Public the way Pieter Levels does — it used Midjourney + Runway to mass-produce ad creative and bought traffic directly. That’s the other solo-founder growth path: spend on traffic and use AI to crush CAC. How to do it: generate 20 ad variants at once with Midjourney, small-budget A/B test, then scale spend on the highest CTR variant.
Limits and risks
- Not a pure AI/tech product: Medvi’s core value is the convenience of GLP-1 drug prescriptions; AI is the ops tool, not the product. It’s not directly comparable to “I built an AI SaaS” — it’s more like “running a traditional business with AI.”
- Heavy regulatory risk: Telehealth prescriptions (especially GLP-1 class) face the possibility of stricter FDA enforcement. Policy changes could wipe out the model overnight.
- Outsourcing dependency risk: Core medical infrastructure is entirely outsourced — if CareValidate or OpenLoop Health pulls back or raises prices, Medvi’s ops break.
- Revenue scale ≠ sustainability: $401M first-year revenue is stunning, but the GLP-1 market is saturating quickly (Hims, Ro, Done, and more competitors).
- The “2 people” boundary is fuzzy: Only 2 official employees, but heavy lifting flows through AI Agents and outsourced providers — actual operational complexity is much higher than it looks.
Why this case matters
As of April 2026, Medvi is the closest real-world case to Sam Altman’s “one-person billion-dollar company” prediction. The key paradigm it reveals:
The competitive edge in the AI era isn’t “a better AI model” — it’s “a leaner organizational architecture.”
A traditional company (Hims & Hers) uses 2,442 people to reach $2.4B in revenue at 5.5% margin. Medvi uses 2 people + AI to reach $401M in revenue at 16.2% margin. The revenue-per-head difference is over 1000x. The solo ceiling is no longer “how much code one person can write” — it’s “how much AI + outsourced infrastructure one person can orchestrate.”