AI Side Hustles Half-true — works only if you do the unspoken work

By the Income Reality Check Editorial Team · Updated 2026-09-02 · 7 min read

A 1-person AI business built in a day with Claude — the part the video doesn’t film

Verdict: Half-true — works only if you do the unspoken work. The stack is real and the build is genuinely fast; what’s missing is a single paying customer and the ad math to get one.

Ishan Sharma’s video opens with a jaw-dropper: Matthew Gallagher, a founder who built a company that made $400 million in 2025 and is projected to hit $1.8 billion in 2026, mostly on the back of AI tools. Then Sharma — a Bengaluru creator with more than 1.7 million YouTube subscribers and his own marketing agency — spends 20 minutes wiring up his own “one-person business” using Claude, Replit, a voice agent, and about six other tools, all in under 24 hours. It’s a slick demo. But by the end, he hasn’t sold a single ₹500 (~$6) session, and the interesting stuff — the part that decides whether any of this makes money — never gets filmed.

What the video actually claims

The pitch is that you can now assemble an entire business with AI acting as the operator, not just the intern. Sharma picks an idea (weekend in-person meetups for lonely 20-somethings in tier-one Indian cities, modeled on a real company he name-checks), then builds it live.

The workflow is dense. He uses a Claude “expand and contract” skill to interrogate the business plan (he says it took 11 minutes of back-and-forth). He builds the landing page on Replit, makes a registration form with Gemini inside Google Forms, generates cafe photos with ChatGPT’s image model, and creates video ads through a Higgsfield MCP server connected to Claude. He wires a voice agent on Retell AI to call everyone who fills the form, connects Razorpay and Stripe payment links, plans WhatsApp follow-ups, and uses an Apollo integration to scrape cafe partnership contacts and email them on a schedule. He buys a domain for about ₹380/year (~$4.50) on Namecheap.

The framing is important. This is a “FREE COURSE,” and the underlying promise — echoing the Gallagher opening — is that one person plus AI can now run what used to take a team. On the raw mechanics, that promise mostly holds up.

What the method actually requires

Here’s where the demo and a business part ways.

Start with the anchor story. Gallagher’s company is Medvi, a U.S. telehealth startup selling GLP-1 weight-loss medication, and Forbes reported it did about $401 million in 2025 sales with a tiny team and a 16.2% margin. Real numbers. But that’s a prescription-drug operation riding one of the hottest consumer-health waves in a decade, with licensed pharmacy fulfillment behind it — not a meetup app someone spun up over a weekend. Using it to introduce a ₹500-per-session cafe business is the classic bait-and-switch of income content: the headline result and the demonstrated method have nothing to do with each other.

Now the tools. None of them are free at scale, and the costs stack:

The voice agent is a good example of how “it works” hides “it costs.” Retell’s own pricing shows the advertised $0.07/minute only covers voice infrastructure; a production setup with a language model, text-to-speech and telephony runs $0.13 to $0.31 per minute all-in, plus $2/month per number. Fine for a demo call. A drag on unit economics once you’re calling hundreds of leads, most of whom — like Sharma himself in the video — say “no, I’m not interested.”

Then there’s the line item the video sets to ₹500/day and moves past: ads. Sharma’s whole funnel depends on Meta ads driving form fills. In India, cost per lead for local service businesses commonly runs ₹100–₹500, and marketers generally say Meta’s algorithm needs roughly ₹15,000–₹30,000/month before it optimizes at all. A ₹500/day budget is ₹15,000/month — the floor, not a growth budget. At even ₹200 per lead, that buys ~75 leads. If a voice agent qualifies a fraction and a fraction of those pay ₹500, you’re collecting a few thousand rupees against ad spend plus tool subscriptions. Do you see the problem?

The U.S. Small Business Administration makes the same point in plainer terms: startups need to budget more for marketing, not less, precisely because nobody knows your brand yet. AI builds the storefront in an afternoon. It does not build demand.

Who actually wins this game

The people who win with this exact playbook are the people who already have distribution. Sharma is one of them — a creator with a seven-figure audience, an existing agency, a Stripe account with live products, and (as he mentions) an “enterprise AI training” list his own voice agent references. When your top-of-funnel is a YouTube channel with millions of views, the ad math the rest of us face barely applies.

The other winners are operators with real domain expertise and capital, like Gallagher — who had a specific, high-margin market and money to spend before AI entered the picture. Sam Altman has floated the idea of a one-person billion-dollar company, and CNBC has covered his broader claim that this is a strong time to start a career with these tools. Notice the pattern: the people making that case are already at the center of enormous audiences and platforms. The tools lower the build cost. They don’t hand you customers.

What you’d realistically earn

The video never states an income figure for the meetup business, which is honest in one sense and slippery in another — the $1.8 billion is doing the emotional work up front. So let’s be concrete about a true beginner running this stack cold, with no audience.

Month one: likely ₹0 in profit and a few thousand rupees out the door on ads, subscriptions, and per-minute agent calls. A local meetup at ₹500/head with groups of three to five is ₹1,500–₹2,500 gross per event, before cafe costs and no-shows. To clear a modest ₹20,000/month profit you’d need to run and fill many sessions consistently, which means solving the two things AI didn’t do for you: getting cheap, reliable leads, and getting cafes to partner. Realistic range for someone starting from zero is roughly ₹0 for the first month or two, then maybe ₹5,000–₹25,000/month if — and only if — the ad funnel and retention actually work after months of tuning. That’s a real small business, not passive income.

One more thing worth flagging for U.S. readers building income-claim content: the FTC has proposed rules requiring sellers of “money-making opportunities” to substantiate earnings claims on request. Opening with a $1.8 billion figure to sell a course sits in exactly the zone regulators are watching.

Who this is (and isn’t) for

This makes sense if you already have an audience or a paying client base, a few hundred dollars a month for tools and ads, and the patience to treat the AI build as day one of a months-long grind. If you enjoy operations — testing creatives, following up leads, closing partnerships — the stack genuinely compresses weeks of setup into a day.

It’s not for someone expecting the video’s implied outcome: press a few buttons, let agents run, watch money arrive. If you have no distribution, no ad budget, and you’re hoping “AI does it all,” you’ll spend on subscriptions and discover the hard part was never the website.

What to remember

The build is real. Sharma isn’t faking the tools, and assembling a landing page, a form, ad creative, a voice agent, and payment links in an afternoon would have been science fiction three years ago. What’s missing is everything after “publish” — the ad economics, the conversion rate, the cafe deals, the months of tuning. AI moved the starting line up. It didn’t move the finish line closer.

Related reading on this site: 9 one-person businesses with more demand than competition and 10 Claude AI side hustles that can pay a full-time income.