A 41-year-old founder in Los Angeles launched what might be the most talked-about AI startup of 2026: a telehealth company called Medvi. He started in September 2024 with $20,000 and no employees. He used ChatGPT and Claude for code and copy, Midjourney for ad creative, and ElevenLabs for customer calls. First-year revenue: $401 million. His only hire was his brother.
Medvi now projects $1.8 billion in 2026 sales, with a 16.2% net profit margin. For comparison, Hims & Hers did roughly $2.35 billion last year with 2,442 employees and a 5.5% margin.
Every founder reading this is thinking: can I do that? Probably not.
What this AI startup actually did right
Matthew Gallagher didn't build a tech company. He built a marketing machine that sits on top of other companies' infrastructure.
The regulated stuff (licensed physicians, prescriptions, pharmacy fulfillment, shipping) was all outsourced to CareValidate and OpenLoop Health. Medvi kept the parts AI is good at: branding, website, paid media, checkout flow, and customer communication.
That's the part worth paying attention to. Gallagher picked a market with massive demand (GLP-1 weight loss drugs), found partners who handled the hard parts, and used AI to run everything else at almost zero headcount.
Can AI really replace a software team?
Malaysian founders keep asking us this. And the answer depends entirely on what you're building.
If your product is basically a front-end for someone else's regulated service, then maybe. Medvi's "product" is a marketing funnel with a checkout page. AI tools can handle that.
But if you're building custom software with business logic, database integrations, payment processing, or anything that needs to work reliably at scale, AI alone falls apart. We use AI coding tools daily at Gotchaa Lab. They make our team faster, not smaller. Big difference.
AI writes code. It doesn't make architecture decisions, debug production failures at 2am, or understand why your Malaysian client needs SST calculations handled differently from GST.
| Task | AI can handle | Still needs humans |
|---|---|---|
| Marketing copy and ads | Yes | Brand strategy |
| Basic website or checkout | Yes | Custom business logic |
| Customer support scripts | Yes | Escalation and empathy |
| Regulatory compliance | No | Always |
| Architecture decisions | No | Always |
| Local context (PDPA, MOH) | No | Always |
The part the hype articles skip
Most coverage of Medvi leaves this out: in late February 2026, the FDA sent warning letters to 30 telehealth companies, including Medvi, for misbranding violations related to compounded GLP-1 drugs. Medvi's website language falsely suggested it was compounding the medications it sold, and the FDA flagged claims implying FDA approval of compounded products. AI can generate thousands of ad variations and write compelling copy. It can't tell you when that copy crosses a regulatory line.
What this means for Malaysian businesses
Malaysia has its own version of this risk. The PDPA governs how you handle customer data, MOH licensing controls health services marketing, and Bank Negara regulates fintech.
If you're a Malaysian founder thinking about building an AI-first company, here's what we'd tell you over coffee:
Copy the structure, not the shortcuts. Medvi's real insight was outsourcing regulated work to licensed partners and keeping the customer relationship. That works in Malaysia too, but you need licensed partners and compliant processes.
Use AI where it actually helps: content, customer support, ad creative, data analysis. We help clients build AI solutions that automate these workflows every week.
Keep humans on compliance. No AI tool will tell you whether your marketing claims comply with Malaysian Advertising Standards Authority guidelines.
And don't confuse revenue with sustainability. An FDA warning letter in year one is a red flag. Building fast without guardrails is how companies get shut down.
Our take on AI startups in Malaysia
The Medvi story is exciting. It's also misleading. AI tools really can compress what used to take 50 people into what 2 people can manage. That part is real.
But it worked because the product is commoditized (compounded GLP-1 drugs), the infrastructure is fully outsourceable, and demand is so high that aggressive marketing alone drives growth.
Most businesses don't look like that. If you're building custom software or a SaaS product, you need people who understand your business logic.
We use AI every day. We're fans. But we've seen what happens when people treat it as a replacement for thinking instead of a tool for thinking better.
Wondering what AI can actually do for your business? Talk to us, we'll give you an honest assessment instead of a hype pitch.
