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

By the Income Reality Check Editorial Team · Updated 2026-06-02 · 7 min read

An AI agent “made” $8,000 a month — what the video skips about who did the work

Verdict: Half-true — works only if you do the unspoken work. The agent scraped the leads and built the sites; a human salesperson and an existing agency turned that into recurring revenue.

In “I Told an AI Agent to Make Me Money. It Did.”, Chris Koerner sits across from Brandon Doyle of getdavid.ai, who claims an AI agent he calls “Zach Morris” spun up a business that now does “a little over 8K a month” in monthly recurring revenue. The agent — they call the software OpenClaw — supposedly found 350 local businesses, built 350 websites for “a handful of cents” each, and mailed out postcards, all from a Mac Mini in Provo, Utah that Brandon texts via iMessage. The number is probably real. But “an AI agent made me money” hides the part that actually generated the cash, and that part is the whole story.

What the video actually claims

Brandon pitched three businesses an agent could launch. One was Facebook Marketplace arbitrage; he abandoned it, so set it aside. The other two he says made money.

The first is a custom bedtime-story service — DreamTales — that emails a personalized story to a parent’s inbox at 7 p.m. nightly for $9 a month. Brandon says the agent built the site, he made one ad, and the business reached “a little over $300” in MRR from roughly 33 to 34 customers after about $400 in Facebook spend. By his own account, it’s “basically break even.”

The second is the headline. Brandon told the agent to scrape Google Business listings across Utah for local trades — landscaping, HVAC, plumbing, roofing — that had no website or a bad one. It returned 350 businesses with names, phone numbers, and emails, then generated 350 websites on cheap “Chinese models” hosted on test URLs (never on live domains). Using a direct-mail API called Lob, it sent each owner a postcard with a QR code linking to their new site. Of 350 postcards, 341 were delivered and about 68 people (20%) scanned. Then Brandon handed the list to “one part-time sales guy” who cold-called everyone. The result: about 17 to 20 customers paying an average of roughly $400 a month, a couple at $1,200, for a total he puts near $8,000 MRR. Token cost for the whole thing? He says $10 to $20, plus about $400 in postcards.

To his credit, Brandon says the quiet part out loud at one point: the agent “did every single thing except they didn’t close the final deal, which was my part-time sales guy or me.”

What the method actually requires

Here’s the math the framing skips. The $8,000 is gross recurring revenue, not profit, and it isn’t payment for websites. It’s payment for marketing retainers — Brandon’s existing agency, David AI, sells “AI-powered marketing services” starting at a few hundred dollars a month, with packages he describes in the $300–$700 range. The free website is the hook. The recurring money is a service his company now has to deliver every month, forever, with humans and software, to keep those 17–20 clients from canceling.

So three costs never appear on screen. The first is the salesperson. Cold-calling 350 businesses to close roughly 17 is real labor, and it isn’t free — U.S. sales reps and telemarketers don’t work for postcard money. The second is fulfillment: every dollar of that MRR carries an obligation to actually run marketing campaigns. The third is the agency itself. Brandon already had the packages, the pricing, the sales process, and a brand. The agent plugged into infrastructure most viewers don’t have.

Then there’s the legal piece the video waves at but doesn’t pin down. Chris warns that mass-texting is “not compliant” and will get you fined. He’s right, and the same caution extends to the cold calls. In the U.S., the FTC’s Telemarketing Sales Rule was amended in April 2024 to cover many business-to-business telemarketing calls, prohibit material misrepresentations, and require detailed call records kept for five years. Pitching small-business owners by phone at scale is a regulated activity, not a growth hack. (U.S. readers, that’s the FTC; if you’re in the U.K. it’s the ICO and Ofcom rules on unsolicited calls, and most English-speaking markets have an equivalent.)

There’s also the load-bearing assumption that an AI agent can reliably do all this unsupervised. The evidence there is mixed. When Anthropic ran Project Vend — letting Claude operate a small office shop — the agent lost money, sold items below cost, hallucinated payment details, and “did not reliably learn from these mistakes.” CNBC has reported that companies are spending big on agentic AI without always knowing what it does, with autonomous systems failing quietly at scale. An agent that scrapes a list and fills a template is on solid ground. An agent trusted with a credit card and a sales funnel is a different animal.

So who actually pocketed the $8,000?

Read the transcript closely and the winner is obvious: an established marketing agency with a product to sell and a person to sell it. The agent compressed the grunt work — lead research, throwaway site generation, addressing and mailing postcards — into cents. That’s genuinely useful, and it’s the real, defensible takeaway here. What it did not do is invent demand, close a sale, or service a client.

This is the pattern across nearly every “AI built my business” video. The AI handles production. A human handles distribution and delivery. Brandon’s own data makes the point twice: the bedtime-story ad that converted best was the one of him talking to a camera, not the AI-generated creatives, and roughly 90% of his website closes came from owners who’d scanned the postcard and then talked to a salesperson. People bought because a person followed up. The agent set the table; it didn’t seat the guests.

Who replicates this? Someone who already runs a service business and wants cheaper lead-gen. Not someone starting from zero.

What you’d realistically earn

If you have no agency, no sales rep, and no service to deliver, copying this nets you a folder of 350 unhosted websites and a stack of postcards. Could you sell some? Maybe. But you’d be cold-calling strangers to pitch a service you’d then have to actually perform, in a market with more than 36 million U.S. small businesses and no shortage of people already emailing them “I can get you to #1 on Google.”

A fairer expectation for a true beginner: weeks of setup, a few hundred dollars in mailing and tooling costs, a learning curve on call compliance and pitching, and — if you stick with it — a handful of small retainers that you then have to earn every month through real work. The bedtime-story business is the honest mirror here. With the agent doing nearly everything, it landed at break-even on $300 MRR. That’s not failure; it’s just the actual shape of a brand-new subscription product before churn, ad tuning, and word of mouth have time to compound. The $8,000 number is real because the agency around it is real — strip that away and the comparable figure for most viewers is far closer to the bedtime-story result than the headline.

Who this is (and isn’t) for

This makes sense if you already sell a service to local businesses and want to cut the cost of prospecting and demo-building. If you can comfortably run a B2B outreach program inside the law, deliver marketing work clients will keep paying for, and absorb a few hundred dollars in mailing costs to test, an agent like this is a legitimate efficiency lever. It does not make sense if you’re hoping the agent itself becomes the business while you sleep. No salesperson, no service, no audience, no budget for postage and ad testing — and the realistic outcome is a pile of unsold websites and a token bill.

What to remember

The agent did real, valuable work, and Brandon isn’t lying about the $8,000. He’s just letting the framing imply that the software earned it. It didn’t. A human salesperson, an existing agency, and a real service did the closing, the delivering, and the keeping. Treat AI agents as the cheapest employee you’ll ever hire for grunt work — and remember that someone still has to run the company.

Related reading on this site: this new AI agent turns you into a one-person company and launch a $10k AI business solo, no code, no funding.