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

By the Income Reality Check Editorial Team · Updated 2026-08-12 · 7 min read

AI marketing agents to grow your startup: the parts the masterclass waves past

Verdict: Half-true — works only if you do the unspoken work. The agent stack is cheap and real; the traffic it produces depends on deliverability, targeting, and compliance work the video treats as a footnote.

Greg Isenberg’s podcast episode “Marketing Agents Masterclass (GROW your startup)” brings back operator Cody Schneider to demo something seductive: AI agents that “get customers on autopilot.” The pitch is that you can point Claude Code or Codex at a stack of about 20 tools, wire up a cron job, and stop worrying about traffic and revenue while your “marketing machine” runs. No income number is promised. Something bigger is — the end of the distribution problem. That part is where the honesty gets thin.

What the video actually claims

The episode walks through two agents in detail. The first is a cold-outbound machine: it monitors LinkedIn posts from influencers in your niche, scrapes everyone who liked or commented (using a scraping API called Apify), runs a “waterfall enrichment” through data brokers like getleads.io, Apollo, and Lead Magic to find emails and phone numbers, validates those emails with MillionVerifier, then blasts cold email through Instantly and LinkedIn DMs through HeyReach or Bot Dog — with an LLM managing the replies and pushing people toward a booked demo.

Schneider is refreshingly specific on cost. He says you can send roughly 10,000 cold emails a month for about $100 in inbox infrastructure (he names Hypertide, Inbox Kit, and Instantly’s ~$97/month tier), and that the full sending setup runs “about $200 to get started.” No course, no gatekeeping — he explicitly tells viewers to DM him instead of buying anything. The second agent is the organic mirror: record a weekly interview with your team, extract insights with Claude Sonnet, and auto-schedule LinkedIn posts across multiple accounts through a tool called Ordinal, feeding the analytics back into the next round.

His framing is that “marketing agents are the new coding agents” — that marketing is now just software, and you shouldn’t pay Anthropic or OpenAI per token when you can build cheap software that only calls an LLM when it actually needs to think.

What the method actually requires

Here’s the thing the demo glides over: an outbound pipeline is only as good as its reply rate, and cold email reply rates have been falling for years. Independent 2025 benchmarks put the B2B average around 3–5%, down from roughly 8.5% in 2019, with platform-wide data from Instantly’s own report landing near 3.4%. Schneider says it himself in the episode — “cold email is getting decimated, reply rates are down, everything is down” because “AI slop is flooding the zone.” He’s right. He’s also selling the tool that helps flood it.

Run the math on his own numbers. Send 10,000 emails at a 3% reply rate and you get 300 replies — but replies aren’t customers. If a healthy fraction are “no,” and meeting-booking rates on cold campaigns sit closer to 1–2% of sends, you’re looking at maybe 100–200 booked calls in a strong month, before you subtract everyone who no-shows. That can absolutely be worth it. But “you don’t have to worry about revenue” is not what those figures describe.

Then there’s the law, which the video handles with a shrug and a “do your own research.” In the U.S., cold email is legal, but the CAN-SPAM Act sets hard rules the FTC enforces: accurate header and “from” information, a subject line that matches the content, clear identification that the message is an ad, a valid physical postal address, and a working opt-out honored within 10 business days. Each violating email can draw a civil penalty of up to $53,088, with no cap on the total — the FTC’s compliance guide spells it out. In August 2024 the FTC hit camera maker Verkada with a $2.95 million CAN-SPAM penalty, the largest in the Act’s history. (U.S. readers, that’s your jurisdiction; the rules elsewhere are stricter.)

Send into the EU and the calculus changes entirely. B2B cold email there leans on the “legitimate interest” basis under Article 6(1)(f) of the GDPR — not a loophole, but a documented balancing test, as the European Commission’s guidance describes. Germany, Italy, and Spain effectively require prior opt-in for B2B outreach. GDPR fines run up to €20 million or 4% of global revenue. A cron job scraping engagers and emailing them across borders doesn’t get to ignore any of that.

And the scraping itself sits on shakier ground than “fully legit.” Schneider calls it white-hat. The reality: LinkedIn’s User Agreement flatly prohibits automated scraping and data collection. In the hiQ Labs v. LinkedIn saga, hiQ won the narrow point that scraping public data isn’t a federal computer-crime violation — but LinkedIn won on breach of contract, and the 2022 settlement entered a $500,000 judgment against hiQ and effectively barred it from scraping the platform. Accounts that automate get restricted or banned. That’s a real operational risk sitting under the whole first agent.

Is this actually cheaper than paying for tokens?

Schneider’s sharpest idea is that you should build software that runs on cheap compute and only “pays the token tax” when a decision genuinely needs an LLM. For a repeatable pipeline — scrape, enrich, validate, send — that’s sound engineering, and CNBC has reported since 2023 on big companies wiring ChatGPT into outreach exactly this way. The savings are real at the production step.

But cheap production is the whole problem the industry now has. When every startup can generate personalized outreach for pennies, the scarce thing isn’t the software — it’s a message a stranger actually wants to open. The agent lowers your cost per email. It does nothing, on its own, to raise your reply rate. Those two facts point in opposite directions, and the video only celebrates the first.

Who actually wins this game

Look closely at who Schneider is describing, and it isn’t a solo beginner. He says the businesses this works for “already know who their target customer is interacting with.” His own social strategy leans on a two-year corpus of posts he knows will go viral because he’s tested them for years. His company, Graft, “forward-deploys software engineers” to build these implementations for “fast-growing companies.” That’s the tell. The winners here are operators who already have a sharp ICP, existing distribution or a personal brand, and enough deal size to make 100 cold conversations pencil out. The agent compresses their labor. It doesn’t manufacture the judgment, the offer, or the audience underneath it.

What you’d realistically earn (or spend)

Nobody promised a dollar figure, so measure this in customers and cost. Budget the honest version: ~$200/month for sending infrastructure, plus per-lead enrichment fees across getleads/Apollo/Lead Magic, plus MillionVerifier, plus a server (Railway or similar), plus your own hours writing offers and babysitting deliverability. Call it $300–$600/month all-in for a modest operation. Against that, B2B customer acquisition costs commonly land anywhere from $300 to over $1,200 per customer depending on segment. If your product’s lifetime value clears a 3:1 ratio over that CAC, outbound agents can be a genuinely good channel. If you’re selling a $19/month app to strangers who never raised their hand, the arithmetic doesn’t close no matter how cheap the tokens are.

Would you rather have 10,000 automated emails or 50 warm ones? For a lot of businesses the honest answer is the second — which is exactly why Schneider’s “engagement signal” targeting is the smartest part of the pitch and the hardest to automate well.

Who this is (and isn’t) for

This makes sense if you run a B2B offer with a four-figure-plus deal size, you already know which creators your buyers follow, you can afford the enrichment and inbox spend, and you have the appetite to manage spam-law compliance in every market you email. It does not make sense if you’re a first-time founder hoping to skip the traffic problem, if your product is low-ticket or B2C, or if you’re in or emailing the EU without a documented legitimate-interest process. For those people, the agent just automates the fast path to a burned domain.

What to remember

The tools are real, the ~$200 price tag is real, and the “marketing is software” instinct is genuinely useful. What’s oversold is the promise that this ends your worry about traffic and revenue. The pipeline the video builds is the easy 20%; the reply rates, the deliverability, the offer, and the law are the 80% it treats as obvious. Do that unspoken work and this is a legitimate growth channel. Skip it and you’ve automated your way to a spam folder.

For more on where AI genuinely helps versus where the pitch runs ahead of reality, see our looks at AI running a full content business and Claude AI side hustles that can pay a full-time income.

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