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

By the Income Reality Check Editorial Team · Updated 2026-10-03 · 7 min read

Ishan Sharma’s AI video agency: the ₹15,000/month pitch, checked

Verdict: Half-true — works only if you do the unspoken work. The AI genuinely makes the videos; it does almost none of the selling, and the selling is the whole business.

In “I Asked ChatGPT dots to Make Me Money!”, creator Ishan Sharma wires up OpenAI’s new Dots agent to a video tool called Artlist and tries to build a marketing agency that films AI ads for local cafes and sells them for ₹15,000 (~$170) a month. He calls it “honestly the laziest way to make money with AI.” The production side is real and cheap. The income is not shown — by the creator’s own admission, day one brought zero paying clients — and the sales, the subscriptions, and the calling rules are the parts the video glides past.

What the video actually claims

The workflow is genuinely clever. Sharma connects ChatGPT’s Dots agent (OpenAI rolled out “Dots” agents at its DevDay in late September 2026, per CNBC) to Artlist via an MCP connection, points it at a nearby cafe’s Google Maps listing, pulls the cafe’s own photos, and generates a UGC-style “day in the life” ad plus a 3D cartoon variant. The clips look convincing — handheld shake, natural light, the works.

Then the agent is supposed to sell. Sharma loads a GitHub “marketing skills” repository so the agent can analyze a business, write cold emails through Apollo.io, and even place phone calls through a tool called Retell AI using a US number he bought. The pitch to each cafe: here’s a free sample video, and if you like it, we’ll make six to ten videos a month for ₹10,000–20,000.

On pricing, he’s specific. A 720p sample costs “about 300–400 rupees” on Artlist; a finished 1080p video “about 1,700 rupees” (~$20). Charge a ₹15,000 retainer for ten videos a month, he says, and “you probably pocket 7 to 8,000 rupees.” His closing promise: if the video gets a thousand likes, he’ll post a day-10 update showing what the business actually earned. In other words, the money is a cliffhanger, not a result.

Does the math actually close?

Run his own numbers and they don’t line up. Ten finished videos at ₹1,700 each is ₹17,000 in generation cost alone — on a ₹15,000 retainer, that’s a loss before you’ve sent a single email. To “pocket 7 to 8,000 rupees” out of ₹15,000, you’d be delivering four or five videos, not ten. The headline retainer and the headline margin describe two different businesses.

And the ₹1,700-per-video figure isn’t the whole bill. Artlist doesn’t sell videos à la carte to a free rider; it’s a subscription. Its AI plans start around $23/month (roughly ₹2,000) for a capped pool of monthly credits, and there’s no free trial anymore. Apollo.io, for the email addresses, runs $49/month on its Basic tier (billed annually) or $65 month-to-month, with credits that expire every month and no rollover. Retell AI charges per minute — a base around $0.07 and a realistic all-in rate near $0.13 once you add an LLM and telephony — plus the number itself. Stack those and you’re carrying roughly ₹6,000–8,000 in fixed monthly cost before you’ve earned a rupee. Your first client’s entire “profit” goes to the tools.

None of this makes the business impossible. It makes it a business — one with overhead, a sales funnel, and a breakeven point — rather than a lazy money printer.

What the method actually requires

Here’s the part the two-hour demo can’t fake: demand. Making a video is now trivial; getting a cafe to pay for one every month is the actual job, and it’s the same cold-outreach grind agencies have always done.

Sharma shows this failing in real time. “No one was really interested in replying to a AI chatbot,” he admits of the Retell calls, so he pivots to email. That tracks with the wider evidence. CNBC reported in April 2026 that AI phone and chat agents are off to a “rocky start,” with roughly one in five customers who used AI service seeing no benefit at all — a failure rate about four times higher than for AI generally. People can tell, and they hang up.

Cold email isn’t a magic fix either. Industry benchmark reports for 2026 put the average B2B cold-email reply rate between roughly 0.5% and 3.7% depending on how you count, with 5% considered top-quartile. Send 500 cold emails and a realistic “interested” count is single digits — before anyone has paid. That’s why marketing budgets are thin at exactly the businesses he’s targeting. The U.S. Small Business Administration notes the average business spends about 1.08% of revenue on advertising and restaurants about 1.93% — a small independent cafe doing, say, ₹15 lakh a year in revenue is working with a few thousand rupees a month for all marketing. A ₹15,000 retainer is a big ask for that buyer.

There’s also a rule the video ignores. In India, promotional calling is governed by TRAI’s telecom commercial-communications framework: commercial callers are meant to register on the DLT platform, use the 140 number series, scrub lists against the Do-Not-Disturb register, and call only between 9 a.m. and 9 p.m. Penalties for violations scale into lakhs per month. Buying a US number to dial Indian cafes doesn’t sidestep that — it arguably compounds it. (Globally, income-claim rules bite too: in the U.S., the FTC banned a company called Air AI from marketing business opportunities in 2026 over deceptive AI earnings claims.)

Who actually wins this game

The people who make this model work aren’t the ones with the slickest AI pipeline. They’re the ones who can sell. Sharma says it himself — he started his own marketing agency in 2021 and spent five years working with large edtech and finance companies. He already knows how to pitch, write a converting email, handle objections, and close a retainer. The agent didn’t give him that; he’s bringing a decade of distribution to a one-day production demo.

So the winners are experienced operators with an existing network, a comfort with rejection, and the patience to send hundreds of outreach messages for a handful of replies. India has, by one directory’s count, more than 8,000 digital marketing agencies already chasing small-business clients. The AI tooling lowers the cost of making the deliverable; it does nothing to reduce that competition for attention.

What you’d realistically earn

Be honest about month one: likely ₹0 in revenue against ₹6,000–8,000 in subscriptions, exactly as the video’s own day-one footage shows. Land one retainer after weeks of outreach and — using his corrected margins, not the ₹7,000–8,000 he quotes against ten videos — you might net a few thousand rupees after tools and generation costs. Three or four steady retainers, several months in, could become a real part-time income of ₹20,000–40,000 a month. That’s a plausible freelance outcome, and it’s roughly in line with what NerdWallet describes for service side hustles, where the average freelancer earns about $28 an hour and income tracks effort and demand, not automation.

What it is not is passive, and it is not “the laziest way to make money with AI.” The creator never claims a figure he actually banked — the whole payoff is deferred to a sequel that depends on the video hitting a thousand likes. Treat any number in that follow-up the same way: as a single result, not a baseline.

Who this is (and isn’t) for

This makes sense for someone who can already sell, has 8–10 hours a week for outreach, can float a few thousand rupees a month in tools before the first client lands, and treats rejection as data. If you’ve run client work before, the AI genuinely removes the expensive, slow part — shooting and editing.

It’s a poor fit if you’re hoping to “set it and forget it.” If cold calling and cold emailing make you wince, if you have no cushion for subscription costs, or if you believe the agent will close deals while you’re “chilling, living life,” this will mostly cost you money and leave you with a folder of unanswered emails.

What to remember

The production half of this pitch is real — AI can now make a cafe a passable ad for the price of a coffee. The business half is the oldest grind there is: finding buyers, surviving rejection, and closing retainers against thousands of competitors. The agent executes; you still have to sell. Judge the method by the day-10 numbers the creator hasn’t posted yet, not the day-one demo that looked effortless.

Related reading on this site: 5 proven ways to make money with AI (no experience) and AI can now run your entire content business — here’s how.

Views and other figures were accurate at the time of review and may have changed since publication.