The $1.8B AI Company With Two Employees: What the Headlines Left Out

You probably saw the headline. Two brothers. AI did almost everything. $1.8 billion in projected sales. The New York Times ran the profile on April 2. Within 48 hours it was on every business podcast, every LinkedIn feed, every group chat where someone wanted to argue about AI replacing employees.

The story is real. The company is real. The numbers — verified by the Times — are real. But the headline is also missing about 60% of the picture. And if you are a contractor, an electrician, a plumber, a landscaper — somebody who actually has customers walking through your door and a name on the side of your truck — the missing 60% matters more than the headline does.

Here is what the NYT story said, what it left out, and what you should actually take from it.

What the Headline Said

The company is called Medvi. It sells compounded GLP-1 weight loss drugs through an online telehealth model. Founder Matthew Gallagher built it in two months with about $20,000 and a stack of AI tools — ChatGPT, Claude, Grok, Midjourney, Runway, ElevenLabs. The AI wrote the code, generated the website copy, produced the ads, ran customer service, and analyzed business performance.

By the end of 2025, Medvi had done $401 million in sales with one employee. He brought his brother Elliot on, hit 250,000 customers, and is on track for $1.8 billion in 2026 — with a 16.2% net profit margin. For comparison, Hims and Hers, with over 2,400 employees, posted a 5.5% margin the same year.

The framing wrote itself: One man, two brothers, a laptop, a billion-dollar company. Sam Altman predicted this in 2024. Here it is, two years early.

What the Headline Left Out

The Times piece ran on April 2. The FDA had already issued Medvi a warning letter on February 20 — six weeks earlier — for misbranding the drugs the company sells. That letter (#721455, public on FDA.gov) said two things mattered:

• Medvi's website displayed compounded semaglutide and tirzepatide products with the "MEDVi" name on the label, suggesting Medvi was the drug compounder. It is not.
• Medvi's site claimed its products had the "same active ingredient as Wegovy and Ozempic" — which the FDA said falsely implied FDA approval, since compounded drugs are not FDA-approved.

The FDA gave Medvi 15 working days to respond and warned that failure to fix the issues could result in seizure or injunction. None of that made it into the NYT profile.

Then came Business Insider. Their April investigation found that Medvi's ads on Meta were largely run through affiliate marketers — and many of those affiliates were running ads under what appeared to be AI-generated fake doctor profiles. One profile, "Dr. Matthew Anderson MD," listed an Angolan phone number and seemed to have previously belonged to a gospel musician. Another, "Dr. Spencer Langford MD," was tied to a clothing store in the Republic of Congo. After Business Insider raised the issue, Medvi's active Meta ad count dropped from over 5,000 to roughly 2,800 in a few days.

Stack on top of that: a class action lawsuit filed in California in March alleging affiliate spam, an FTC investigation request from the National Consumers League naming Medvi as one of six telehealth companies of concern, and a 2025 Futurism report that the "before-and-after" weight loss photos on Medvi's site appeared to be AI-altered images of real people lifted from across the web.

Gallagher has a defense — he says the FDA letter was about a different domain (medvi.io) operated by an affiliate, not his main site (medvi.org), and that fake-doctor ads are an affiliate problem he is now policing. Maybe that holds up. Maybe it does not. Either way, that is not a billion-dollar company built by two brothers and a laptop. That is a billion-dollar marketing operation built on top of someone else's drug compounding, someone else's pharmacy network, someone else's clinical compliance, and an affiliate army the founder admits he is not fully monitoring.

The Real Lesson Most People Will Miss

The takeaway is not "AI is bad." The takeaway is not "Medvi is a scam." Both of those are lazy. The real lesson is sharper and more useful, and it applies directly to anyone running a service business with a real customer base:

Medvi scaled customer acquisition faster than it scaled compliance, oversight, and quality control. The AI did not break — the AI did exactly what it was told. The cracks were already in the foundation. AI just made them visible faster, to more people, with more money on the table when the FDA, the FTC, and Business Insider showed up.

For a contractor, that should land hard. You do not get a do-over when your name is on the truck. You can't ghost the FDA the way an affiliate can ghost a Meta ad account. You can't spin up a new LLC when a homeowner posts a one-star review with photos of a botched job. The customer who fires you tells fifteen other homeowners in your zip code. That is the world a real operator lives in — and it is the world AI either helps you serve better or helps you wreck faster.

What This Should Change About How You Think About AI

The right takeaway from the Medvi story is not to dismiss AI. The right takeaway is to be specific about where you let AI run and where you keep a human firmly in the loop. The split is not complicated:

Use AI for the parts that should run on rules.

Invoice intake. Estimate templates. Schedule reminders. Bilingual timesheet collection. Crew dispatch confirmations. Follow-up emails to leads who went quiet. These are jobs that should produce the same output every time, on time, without you. AI is excellent at them. You should not be doing them at 9pm on Sunday. Most contractors I work with get 8-15 hours a week back from this category alone.

Keep a human in the loop for anything that touches your reputation, your customer relationships, or a regulator.

Customer service that has any chance of escalating. Final invoices and change orders. Anything you put your license number on. Marketing copy that runs at scale. Reviews and testimonials. AI can draft, summarize, and prepare — but a person needs to look at it before it goes out the door with your name attached. That is not a limitation of AI. That is a feature of running a real business.

Be honest about what your foundation is.

If you are doing good work, taking care of your customers, and treating your crew right, AI is going to make all of that go further and faster. You should be excited. If you are cutting corners, dodging compliance, or hoping nobody notices the cracks — AI is going to make all of that go further and faster too. That should scare you, not excite you.

The Bottom Line for Real Operators

The headline version of the Medvi story is a fairy tale: two guys, some software, a billion dollars. The full version is more useful, and quieter. It is a story about leverage being neutral. The same tools that let a thoughtful operator give themselves their evenings back let a less careful operator scale a regulatory disaster faster than oversight can catch it.

You don't have a billion-dollar drug compounding affiliate network to worry about. But you do have a name on the truck, a license on the wall, a crew that depends on you, and customers who are going to tell their neighbors how the job went. That is your foundation. Use AI to protect it and free up your time. Don't use AI to fake it.

Margin matters more than money. The point of automating your business is not to grow at any cost. The point is to build something that runs without grinding you down — so the time you get back actually goes to the people you built it for in the first place.