I Broke Down How One Person Used Claude to Build a $401M Business. Here Is the Blueprint.

One Person, $20,000, and Claude. Here Is How Medvi Got to $3 Million a Day.

In September 2024, Matthew Gallagher launched a telehealth company from his living room in Los Angeles with $20,000, no employees, and a stack of AI tools.

By the end of 2025, Medvi had recorded $401 million in sales, served over 250,000 customers, and was generating roughly $3 million a day. It currently has two employees: Matthew and his brother Elliot, who joined later to help with communications.

Sam Altman had been betting with tech CEO friends that a one-person billion-dollar company would emerge. When Medvi appeared, he emailed to say he had won the bet.

"It's not an AI company," Gallagher told the New York Times. "But I did it with AI."

The tools he used: ChatGPT, Claude, and Grok for code, copy, and customer service. ElevenLabs for AI voice. Midjourney and Runway for images and video. A handful of infrastructure partners handling the regulated medical side: doctors, pharmacies, shipping, and compliance.

This article is about the blueprint underneath that story. Specifically the parts that work for a standard one-person e-commerce business using Claude, and the parts that require an honest look before you try to replicate them.

What Medvi Actually Is

Most coverage of Medvi frames it as an e-commerce business. It is not exactly that.

Medvi is a telehealth platform operating as a middleman for GLP-1 weight-loss drugs. Gallagher outsourced the expensive, regulated infrastructure to CareValidate and OpenLoop Health, the companies that handle doctors, prescriptions, pharmacies, and legal compliance. What he built on top of that with AI was the customer acquisition layer: the website, the marketing, the copy, the ads, the customer service automation.

The specific vertical, prescription telehealth and GLP-1 drugs, is not something most people can replicate without the same infrastructure partnerships and regulatory knowledge. That is the honest caveat on this story.

What is replicable is the approach: using Claude to do the research, product page writing, and ad strategy that would normally require a team. Applied to a standard physical product e-commerce business, that approach is accessible to anyone.

One more thing worth naming before the blueprint: Medvi has faced serious allegations about its ad practices. Business Insider and other outlets have reported that Medvi ran thousands of ads using accounts claiming to be doctors, which investigators allege were fake. The website reportedly uses AI-generated images of people who do not exist. Gallagher has disputed some of these characterizations. The full picture is contested. What is not contested is that the underlying business approach, AI-assisted research, AI-generated copy, AI-generated ad creative, is the actual mechanism. The ethical line on how you use that mechanism is yours to draw.

What This Actually Costs to Start

Before the three steps, the honest budget breakdown a reader needs before committing a weekend to this.

Minimum realistic budget for a first test:

Realistic total for a first legitimate test: approximately $1,600 to $2,700.

This is not a $0 business. The Medvi comparison is misleading on cost because Gallagher was not holding inventory or running physical goods logistics. He was running a telehealth platform with outsourced medical infrastructure and digital delivery. Physical product e-commerce has inventory, shipping, and return costs that a telehealth middleman does not.

Know the floor before you start.

Step 1: Find a Product With Claude

The first job is market research. Finding something with real demand, reasonable margins, and a problem worth solving that is not already dominated by companies with eight-figure marketing budgets.

The Claude approach to this is not a single search. It is a structured research session using Claude's desktop app with the Cowork feature.

What Cowork is and how to access it:

Cowork is a feature inside the Claude desktop app, not the browser version. Download the Claude desktop app from claude.ai on Mac or Windows. Once installed, open it and you will see a Cowork tab alongside the regular chat. Cowork lets you create persistent projects where Claude retains memory, files, and instructions across multiple sessions. For this use case, you create a Product Finder project, paste your research instructions once as the project brief, and Claude carries that context into every subsequent session without you re-explaining it. This is the key difference from the browser version: context that persists, not a conversation that resets.

The research prompt covers five criteria: passionate market, good profit margins, ships in a standard box, consistent demand (not seasonal or trending), and real pain rather than nice-to-have.

Step 1 Setup

Create a new project in Claude Cowork called Product FinderPaste this instruction as the project brief:You are an expert e-commerce product researcher. Your job is to find validated product opportunities in established markets. For any market I ask you to evaluate, assess: Is this a passionate market with buyers who actively seek solutions? Are margins above 40%? Does it ship in a standard box? Is demand consistent year-round, not driven by trends? Is there a genuine pain point rather than a convenience? Rank any opportunities you identify using these five criteria. Show your reasoning.Then start the research with this prompt:Scan the e-commerce market and identify three validated niches that meet all five criteria above. For each, find two to three specific product ideas within that niche. Do not suggest anything that requires a prescription, medical claim, or regulatory approval. Focus on physical products with clear demand signals.

Claude will typically take ten to fifteen minutes on a request this broad and come back with a ranked list of markets and product ideas within each.

The example from this blueprint: Claude identified pet wellness as a strong market and within it, the senior dog hip support sling, a mobility harness for older dogs that struggle with stairs and day-to-day movement. Five thousand purchases per month on a single Amazon listing. Multiple competing brands scaling ad spend.

Validating the product before committing:

Two checks before moving forward.

First, Amazon. Search the product type and look at purchase volumes. Multiple listings getting consistent monthly sales is the signal. One listing doing huge volume is not the signal because it means one dominant player. Several listings doing moderate volume means the market has room.

Second, Facebook Ads Library. Go to facebook.com/ads/library and search the product category. If competitors are running dozens of ads and have been for months, they are profitable. A competitor with 200 active ads is scaling. That is the confirmation you need.

Claude prompt for Amazon and ads validation:

I am considering selling [product type]. Here are my Amazon search results [paste or describe findings]. Here is what I found in the Facebook Ads Library [describe what you saw].Based on this: is this product validated? What does the competition landscape tell me about my chances as a new entrant? What angle should I take to differentiate from what is already running?

How to Source the Product You Found

The gap between "Claude identified a winning product" and "I have units to sell" is supplier sourcing. Claude cannot close this gap. Here is the manual process.

Where to find suppliers:

Alibaba (alibaba.com) is the most common starting point for physical products manufactured in China or Southeast Asia. Search the product name, filter by "Verified Supplier" and "Trade Assurance," and contact five to ten suppliers simultaneously. Do not commit to a supplier based on their product listing alone.

For US-based suppliers with faster lead times and lower minimum orders, Faire (faire.com) and Thomas Net (thomasnet.com) are alternatives. Margins are lower but you can receive inventory in days rather than weeks.

What to ask suppliers:

Sample request email (paste into Claude and adapt for your product):We are a new e-commerce brand launching a [product name] in the US market. We are comparing suppliers and would like to order samples before placing a production order. Please send: your current unit price at quantities of 50, 100, and 500 units. Minimum order quantity for a first order. Sample availability and sample cost. Lead time from order to shipping. Whether you offer private labeling or custom packaging. Photos of your current inventory for this product.

What to look for in the response:

Suppliers who respond quickly, speak clearly about MOQ and pricing, and offer Trade Assurance through Alibaba are lower risk than those who push you to communicate off-platform or cannot give clear pricing. Samples should look and feel like what you saw in the listing. If the sample disappoints, do not place a production order hoping it gets better.

Lead time reality:

Air freight from China: 7 to 14 days, higher cost. Sea freight: 25 to 40 days, lower cost. For a first test, air freight is worth the premium. You want to be able to iterate on product and packaging quickly, not wait six weeks to see whether the first batch sells.

Step 2: Build the Store With Claude

Once the product is confirmed, the store needs a product page. This is where Claude does something that would have previously required a copywriter, a designer, and several days.

The approach: feed the Shopify product page JSON into Claude, give it your competitor's best-performing product page as a reference, and let it rewrite the copy and regenerate the page structure.

Before the JSON step: set up the store from scratch if you do not have one.

Step 1: Go to shopify.com and start the free trial (no credit card required)Step 2: Choose the Basic plan at $39 per month when prompted — this is all you need to launchStep 3: In the Shopify dashboard, click Online Store → Themes → Browse Free ThemesStep 4: Install Dawn (Shopify's default free theme) — it is clean and fully compatible with the JSON editing approach belowStep 5: Go to Products → Add Product and create your first product listing manually: title, description, price, and at least one imageStep 6: Go to Settings → Payments and connect Shopify Payments (requires a bank account and government ID for verification)Step 7: Go to Settings → Domains and either buy a domain through Shopify ($15 per year) or connect one you already own

Once the store exists and the product is in the system, the JSON editing step below takes five minutes. Without the product in the system first, there is no product template to edit.

Step 2 Setup

In Shopify, go to Online Store → Themes → Edit CodeFind your product template file (product.json or product-template.liquid)Copy the entire contents of the filePaste into Claude with this prompt:I am going to give you the JSON code for a Shopify product page. Read it and understand the structure completely before doing anything else. Confirm when you understand it.Then give Claude a second prompt:Now I want you to rewrite this product page for [your product]. Here is the URL of my best-performing competitor: [URL]. Scan their product page and extract: their main headline, primary benefit statement, how they describe the problem, key features they highlight, the social proof structure they use, and their guarantee or return policy language.Then rewrite my product page JSON using this structure, adapted for my product and brand. Replace all their copy with original copy for my brand. Return the complete updated JSON file.Once you have the new JSON, paste it back into Shopify and click Save.

The result: a fully written product page with benefit-led headline, feature descriptions, social proof structure, FAQ, and return policy, generated in under five minutes. The only manual steps are swapping in the product images and replacing any competitor brand name references Claude may have retained.

The honest caveat on the page build: Claude will occasionally miss edge cases like custom metafields or app-injected sections. Review the output before going live and check the live page for any rendering errors.

Step 3: Create Ads With Claude and Higgsfield

Getting customers requires ads. The approach here has three stages: find what is already working, ask Claude to improve it, then generate the creative with Higgsfield.

Finding what works:

Go to the Facebook Ads Library and search for your product category. Sort by date to find the oldest running ads. An ad that has been running for three months is profitable. Download the image from two or three of the longest-running ads from competitors.

Claude for the creative brief:

I am attaching my competitor's top-performing ad image. I want to create a similar ad for my version of this product. Do not copy it directly. I need:1. A new headline that is more emotionally resonant than theirs2. A sub-headline that reinforces the primary emotion3. A different context or setting for the image that they have not used4. A Higgsfield image generation prompt that will produce a professional ad image in a similar style to the reference, using my new creative directionReference the competitor's ad structure but make everything original.

Claude will return a full creative brief including the Higgsfield prompt.

Generating the image in Higgsfield:

Higgsfield is a purpose-built image generation tool for ad creative. Take the prompt Claude generated, paste it into Higgsfield, upload the competitor ad as a reference image for style guidance, set it to generate four variations, and run it. The output is a production-ready image ad that matches the format of what is already working while being original creative.

Setting up your Facebook ads account before you can run anything:

Meta requires a Business Page and an Ad Account before you can run any paid ads. This takes 24 to 48 hours to fully activate.

Step 1: Go to business.facebook.com and create a Meta Business Suite accountStep 2: Create a Facebook Business Page for your brand (even a minimal one with logo and description)Step 3: Inside Meta Business Suite, go to Ads → Create Ad AccountStep 4: Add a payment method and verify your identity (Meta requires this before your first campaign)Step 5: If your product is in the health or wellness category, check whether it triggers Meta's Special Ad Category requirements — pet products generally do not, but supplement or medical claims will restrict your targeting options

Do not start building the campaign until the ad account is verified and the payment method is confirmed. First campaigns from new accounts sometimes trigger a brief review period. Budget an extra day or two for this.

Testing budget: $30 to $50 per day per ad creative. The goal in week one is not profitability. It is finding which creative direction generates the lowest cost per click with the highest click-through rate. Once you identify the winning angle, scale the budget behind it.

How to read your test results:

The numbers that matter and what they should look like for a physical pet product on cold Facebook traffic.

When to kill an ad: if you have spent three times the product's price and made zero sales, the creative is not working for that audience. Do not increase budget hoping it improves. Change the creative or the audience first.

When to scale: if CPA is comfortably below your margin per unit and the CTR is above 1.5%, double the daily budget every three to five days rather than multiplying it immediately. Rapid budget increases often reset the Meta algorithm's learning phase and degrade performance temporarily.

What the Blueprint Requires to Actually Work

Three things that are not in the transcript but matter for whether this works.

A product with a real margin. The product research step is only useful if the product you find can actually be sold at a profit. For physical e-commerce, that means 40% gross margin minimum before ad spend. The senior dog mobility harness example works because the product has a clear pain point, consistent demand, and room for a 3x to 4x markup on cost of goods. A trendy gadget with a 15% margin cannot absorb the cost of Facebook ads and be profitable.

Infrastructure you control. Medvi's version of this worked at scale because Gallagher outsourced the things he could not do with AI: medical compliance, pharmacy relationships, doctor networks. For standard e-commerce, the equivalent is a reliable supplier, a fulfillment partner, and clean logistics. Claude can write the ads. It cannot ship the product.

Ethical lines in the creative. The Medvi controversy around deepfake images and fake doctor accounts is worth taking seriously as you build this. AI-generated product imagery showing real-looking results is in a grey area that regulators are actively tightening, especially for health and wellness products. The legitimate version of this approach: use AI to generate concept imagery and ad creative that is clearly stylized, not photorealistic depictions of results. Use real reviews, real customers, real outcomes once you have them. The speed advantage of AI creative does not require fabricating the evidence base.

What Claude Does That a Team Cannot Do Faster

The specific advantage is not any individual task. It is the speed of iteration across all three stages simultaneously.

Finding a product, validating it, writing the page, and producing the first ad used to take weeks and multiple contractors. With Claude, the same workflow takes one focused day.

That compression does not just save time. It changes the economics of experimentation. Testing a new product direction is no longer a three-week commitment. It is a one-day test. The business model that emerges from that is fundamentally different from one where each new product requires a full team and a full production cycle.

Medvi is the extreme version of what that compression can produce when applied to the right opportunity. The blueprint is real. The tools are accessible. The opportunity it points toward is in finding the specific vertical where your own expertise and a Claude-assisted execution layer can move faster than anyone with a larger team.

This article is financial commentary, not financial advice. Do your own research before making any business or investment decision.

Follow @damidefi on X for daily Claude AI tools, crypto analysis, and the full journey to 100K. Bookmark this. Share it with one person who has been thinking about starting an online business and has not yet touched a single AI tool to do it.