AI agents for local businesses are being pitched everywhere right now, usually next to a big income claim and a course to buy. What an owner is actually paying for is far more ordinary: an agent that answers missed calls, trims wasted ad spend, or follows up on stale estimates. The sensible way to buy one is to start narrow, give it read-only access, keep a human approving anything customer-facing, name one person accountable, and keep a kill switch within reach.
The "perfect resume" problem with AI agent claims
Scroll YouTube this week and the pattern is hard to miss. Four tracked channels (Corey Ganim, Jono Catliff, Automate AI Consulting and Tim Chao) are all pitching AI agent services to local businesses, and the loudest income claims sit right beside a guide or course upsell. One video is titled around "$730K Earned." The write-up we could find describes that as self-reported revenue from a Google Ads account, not earnings from selling AI agents.
Treat those numbers like a job applicant's resume that looks impossibly perfect. In a sweep of Reddit, TikTok, X and the web for the same period, we found no audited agency profit-and-loss statement, no churn figure and no client-count breakdown. Several creators close their videos with "comment MONEY" or "comment biz" to get a free guide. That is a lead funnel, and it tells you how the claim is being used, not whether it holds.
What owners actually buy
Owners don't buy "AI agents". They buy answered calls, booked jobs and recovered revenue. The offers that recur are boring and specific:
A missed-call front desk that answers, qualifies the lead and books the appointment.
Ad-leak audits that flag wasted search terms, such as a plumber paying for searches like "plumber jobs near me" where "jobs" means someone looking for work.
Lead and estimate follow-up that reaches out to quotes sitting in an inbox.
Back-office help with files, invoices and paperwork.
The best practical clip in the sweep, from @pro.glitch on TikTok, makes the same point from the other side. A cleaning company with 140 clients tried eight AI tools and still told a customer "we clean on Sundays." The fix was a call to map the workflow before touching any tool.
Think of an agent as a digital super intern
The framing that held up best in the episode is the digital super intern: someone who works the night shift cheaply, and who will make mistakes unless you hand over a narrow job and a strict standard operating procedure. A roofing crew up on a roof can't hear the phone. Every missed call is a homeowner with a leak who calls the next roofer on Google. An agent that does only that one job, answering, taking the details and booking, is easy to judge and easy to supervise.
One case study, with a caveat
One case study in the episode's sources describes an agent called DougBot that a drilling company used to tidy a messy Dropbox account, reorganizing about 1,200 files across roughly 220 folders without breaking links. We could not find independent reporting on it, so read it as one builder's account rather than a benchmark. What it illustrates is the category: tedious, well-bounded back-office work.
What it costs
Pricing claims cluster in a narrow band. Guides and creators quote a setup fee of about $1,000 to $5,000 plus a monthly retainer of about $500 to $2,500. On The Koerner Office Podcast, a guest describes roughly $2,000 for setup and $500 a month for a customer-support agent model. Nearly all of this comes from agencies and course sellers, so read it as asking prices, not proof of what clients pay.
There is also a ceiling. AI receptionists are sold directly to small businesses for roughly $25 to $350 a month, so an agency charging $500 or more has to justify the markup with custom integration, local trust and ongoing management. Tooling and API costs, reported at $300 to $800 a month in one guide, are often left out of quotes.
Bottleneck sessions and the value ledger
Owners are notoriously bad at spotting their own inefficiencies because they sit too close to the work. The approach the sources describe starts with a bottleneck session: find one time-consuming task and fix only that. Treat the agent like a managed digital employee on a monthly retainer, and track it with a value ledger, a weekly report of the human hours and dollars the agent saved. Agency-side writers argue this matters for retention: a working automation becomes invisible, and after a few quiet months the retainer starts to look like a line item with no job.
The four guardrails to demand
Read-only first. Let the agent find the leaks in an ad account, but a human clicks approve on the changes. Think of it as giving the intern a highlighter but not the pen.
Human review of anything customer-facing. A drafted estimate gets checked line by line before the client sees it, because a model can confidently invent a discount that costs you real money.
One named owner. A specific person is accountable for the agent's output.
A kill switch. If the agent goes off the rails you can turn it off instantly, without a phone call to the vendor.
The question to take away
If cheap agents make instant follow-up the baseline for every local business, great service stops being a competitive advantage. The question worth asking is what sets your business apart once everyone answers the phone at midnight.
FAQ
Do AI agents for local businesses really earn $1M in 30 days?
We found no independent confirmation of that claim, and nearly all income claims in this window pair with a paid guide or course. Treat them as marketing, not evidence.
How much does an AI agent for a small business cost?
Quoted setup runs about $1,000 to $5,000 with a $500 to $2,500 monthly retainer, while AI receptionists sold directly cost roughly $25 to $350 a month. Ask what tooling and API costs are included.
What should I check before letting an AI agent touch my data?
Ask for read-only access first, human approval on anything customer-facing, one named owner and a kill switch you can use yourself.
What is the best first job for a small business AI agent?
A narrow, easy-to-judge task such as answering missed calls and booking appointments, or cleaning up wasted ad spend, rather than a broad "run my business" brief.
Companion episode:
Want more practical AI breakdowns for solo builders and small teams? Get the free playbooks at joebuildsai.com.