350 Local Businesses, 350 Postcards, $400/mo Each
Brandon Doyle pointed an agent on a Mac Mini at 350 Utah trade businesses with no site or a bad one, built each a test site, then mailed a QR-code postcard. About 20% of owners scanned, 17 to 20 signed, average $400 a month, roughly $8,000 MRR. Every step of that chain can be priced.
📜 Table of Contents
What the chain looks like
The source is a public interview on The Koerner Office podcast. Brandon Doyle self-reported the numbers, and nobody audited them. But every step in the chain can be priced independently, so we priced it and looked at where automation stops.
His setup: an agent named Zach Morris running on a Mac Mini, taking jobs over iMessage. It scraped 350 Utah businesses in landscaping, HVAC, plumbing and roofing, filtered the ones with no site or a bad one, built each a live test site for a few cents, mailed 350 postcards with a QR code through the Lob API, and handed the scans to a part-time salesperson who called and closed them on a monthly fee.
Result: about 20% scanned, 17 to 20 signed, average around $400 a month, roughly $8,000 MRR.
Pricing each step
The list. Google Places API Text Search (Pro) costs $32 per 1,000 requests with 5,000 free per month. Scraping 350 businesses across a state's worth of cities, at two or three pages per keyword, lands inside the free tier. Overages are a few dollars.
The sites. 350 static pages rendered from a template, hosted on Cloudflare Pages or Vercel free tiers. Marginal cost near zero. The real cost is model calls: a few cents per site.
The postcards. This is the only hard cash out. Lob's published rate effective 1 October 2026: a 4x6 postcard at First Class is $0.545 per piece on the Developer plan, $0.905 at the entry tier, print and postage included.
350 x $0.545 = $190.75. At the entry tier, $316.75.
Sales. He did not disclose the commission. At half of first month's fee, 20 clients x $200 = $4,000 one-off, then pure monthly.
So round one costs about $190 in postage plus a few hundred in model calls, and returns $8,000 of monthly revenue. That ratio is the attractive part.
How far automation goes
| Step | Unattended? | Note |
|---|---|---|
| Scrape the list | Yes | Cron by city and trade keyword |
| Detect a bad site | Yes | Load time, mobile usability, missing call button |
| Build the sites | Yes | Template plus model-generated copy |
| Generate QR cards | Yes | Straight into the print API |
| Mail them | Yes | Batch order from an address list |
| Follow up after a scan | Partial | Automated SMS works, but converts worse than a call |
| Close the deal | No | Local owners trust a person, not an email |
That last row is the whole story. The first six steps run overnight. The seventh needs a human voice. In Doyle's version, that voice was a part-timer, not the agent.
Running it in four weeks
Week one: list and verify only. Do not build, do not mail. Scrape 100 businesses and hand-check 20 to see whether your "bad site" filter is right. If this step is wrong, everything downstream is wasted.
Week two: 20 sites, 20 cards. Not 350. Twenty cards cost about $11 and tell you what scan rate looks like. If it comes in under 5%, the problem is the card or the audience, not the volume.
Week three: add the phone. People who scanned answer calls at a much higher rate than cold leads. Sort the scan list by time and call same-day scans the same day.
Week four: measure renewal. Monthly fee businesses live or die in month two. How many of the first 20 stay decides whether this is worth scaling.
Two things that bite
Address accuracy. A misprinted postcard is money burned. Lob charges separately for address verification (from $25 per 1,000 US addresses). Run it before mailing and do not skip the $25.
Local compliance. US direct mail rules vary by state, and regulated industries add more. Landscaping and plumbing are fine; do not copy this onto financial or medical clients.
My read
Of the AI-earning cases I have looked at, this one has the clearest chain and the most checkable cost. The reason is simple: it never tried to automate trust, only labour.
Its ceiling is equally clear. $8,000 a month comes from one person making 20 calls to land 20 clients. Scale to 200 clients and you need a sales team, and the marginal cost comes back.
The minimum viable test is 20 postcards, about $11 in postage, and one week. That is cheap enough to be worth running.
Written by our editorial team; tools listed here are tested or verified against public sources. Links point to official sites or GitHub repos for reference only — no paid placements.