Everyone on X is selling AI agents to other AI people.
Same buyers, same pitch, same inbox with 40 identical cold DMs in it.
Meanwhile there are roughly 112,000 plumbing and HVAC companies in the US averaging $2.5 million a year, and only 14% of their owners say AI is part of how they run. That gap is the whole opportunity.
Greg Isenberg just dropped an episode on exactly this: how to sell AI to local businesses that actually have money. His guest is James Dickerson, his business partner and the guy behind The Boring Marketer. James is not a guru with a course. He owns Diesel Dudes, a mobile diesel repair company in Charlotte, North Carolina, and he grew it with the same playbook he lays out in the episode.
I pulled the whole thing apart: the niches, the offer, the outreach system, and the pricing math that makes a plumbing company owner say yes.
The guy teaching this fixes semi-trucks for a living
James spent years running agencies for local businesses. Performance ads, SEO, social, all of it.
Then he went and bought his own boring business. Diesel Dudes dispatches mechanics to semi-trucks broken down on the side of the road around Charlotte. Costly problem, urgent need, high ticket. He grew it with ads, funnels, and local SEO. The exact stuff he now sells.
In the episode he demos the workflow live. He types "find me plumbers in Dallas worth pitching" into his own tool, and in about a minute it scrapes Google Maps, Google, and Yelp and comes back with 100 plumbers. It scores them on review count, review velocity, technician headcount, and estimated revenue, because he only wants trades businesses doing over $2 million a year.
Then his agents crawl their ads and websites, find the leaks in their customer funnel, and draft a pitch in the prospect's own brand. About 15 minutes from search to a dialed proposal with real work attached.
Compare that to the old way: paying for list-building credits and clicking through websites one by one. (I did that job once. No shade, but I don't miss it.)
The market: $2M to $20M businesses nobody useful is pitching
The sweet spot James targets is local businesses doing $2 million to $20 million a year.
They can afford you. They quietly suspect they're behind on AI. And the pitches they currently get are cold emails asking for a demo call in the first sentence.
The numbers from the episode, for plumbing and HVAC alone:
- ~112,000 companies in the US
- ~$2.5 million a year in average revenue
- 14% of owners say AI is genuinely part of their operation
- $129 for a single lead in some of these markets
And plumbing is the crowded end, because that's where the private equity rollups are happening. James rattles off niches with real demand and weak marketing: commercial hood cleaning (restaurants are legally required to do it), dental implants, tree services, septic, chimney sweeping, gutter guards, backyard sport courts (huge in the Carolinas right now), backup generators in the Northeast.
The filter is simple. High ticket, urgent need, over $2M a year. Skip lawn care and restaurants. A plumbing job runs thousands of dollars, and nobody waits a week on a ceiling leak. James paid $2,000 the same day his kid did something unspeakable to the shower.
→ Pick the niche where you have a home-field advantage. If I lived in Boston it would be snow removal, not sport courts.
(I wrote up a bigger list of boring business ideas that make money if you want to go hunting.)
Never lead with the word AI
This is my favorite part of the episode, because James has the scar tissue to prove it.
He tested an AI voice receptionist at Diesel Dudes. The result was a pile of voicemails from truckers saying "is this a dumb AI thing?" and hanging up.
These owners are not anti-technology. They just don't buy "AI transformation". Half of them hear "agent" and think real estate. And a chunk of mainstream America actively dislikes AI as a brand word.
So you never sell AI automation. You sell more phone calls and more booked jobs.
"I can deploy an AI agent into your company" gets you deleted. "Your ads send people to a page where they can't book, and it's costing you about $2,600 a month" gets you a reply. Same technology underneath, but only one of them speaks revenue.
The offer: funnels, not AI transformation
Here's the wedge. These businesses are already paying for traffic. The leak is everything that happens after the click.
James audited five Dallas plumbers, all rated 4.8 or 4.9 on Google, all actively running Google ads the week he looked:
- 4 out of 5 landing pages didn't repeat the promise made in the ad
- 3 out of 5 didn't show their Google rating anywhere (hundreds of reviews, invisible)
- 4 out of 5 made you scroll three full screens before you could book a service
These are the good companies in a major metro. The low-hanging fruit is staring at you.
Agencies still charge $20,000 for a good funnel. With a coding agent you can build one in days, on brand, at quality that didn't exist in the drag-and-drop era. That price gap is your margin.
The funnel menu James walks through for an HVAC company:
- A free system check as the ad hook
- A $79 tune-up as the tripwire offer (the leading HVAC companies in his city all run one; 80 to 90% of the rest run nothing)
- Instant pricing, because "call us to find out" loses the impatient buyer
- A "repair or replace?" quiz that lands on a video of the owner giving honest advice. Those replacement installs run $20,000 to $30,000, financed. Trust wins them.
- Follow-up sequences behind all of it, SMS and email firing on their own (this breakdown of business automations covers how those sequences work)
Each funnel gets its own ad campaign. Searching "furnace installation" and searching "AC tune-up" are different buyers, and most companies dump both onto the same tired homepage.
A tripwire, if the term is new: a cheap entry offer that recoups your ad spend. One $79 tune-up buyer liquidates the clicks, and you keep the 15 email addresses that came with them.
Getting clients: the Dream 100 and a brass fitting in a box
The acquisition strategy is old direct response with new tooling bolted on.
Chet Holmes wrote about it in The Ultimate Sales Machine. He studied a list of 2,200 advertisers and found that 167 of them bought 95% of the ads. Everyone chased those 167. So Holmes built relationships with the ones who weren't buying yet, showed up with value every week, and by month six he closed 28 clients in a single month.
That's the Dream 100: pick the 100 companies you'd love to work with and become unignorable to exactly them.
The 2026 version, per James:
- Have your agents research all 100 to the nth degree. Reviews, websites, owners, how their customers talk on Reddit.
- Be the advisor, not the vendor (that's Jay Abraham). Add free value weekly. No demo asks.
- Run Meta ads about helping plumbers in a 25-mile radius. The algorithm finds them. You don't need an audience.
- Send lumpy mail (that's Dan Kennedy). James's example: a box with a brass fitting and a letter that says "you fix plumbing leaks, I fix the leaks in your marketing", linking to a page you already rebuilt for them.
- One trade, one city. You don't need to be the authority. You need to be an authority to 100 specific people.
Notice what's missing: blasting 10,000 cold emails and quitting after two weeks. This is a six-month game. The AI does the research grunt work; the patience is still on you.
(If the agency model itself is what pulls you, I broke down a full course on how to start an AI services business in an earlier post.)
The pricing math that closes an owner
A $5 million plumbing company spends $150,000 to $250,000 a year on marketing. Keep that number in your head when you're wondering what to charge.
The pitch math from the episode:
They're buying 100 leads a month at $129 each. Call it $13,000 a month. If the landing page loses 20% of those leads, that's $2,600 a month going nowhere. And one average plumbing job at $1,714 pays for 13 of those leads on its own.
Sell them the $2,600 a month they're already losing, on their own numbers. The funnel is just how you get it back.
Charge like an agency, run on software costs. An install fee for building the funnel, then a monthly retainer for testing, tweaks, and reporting the call volume. Reporting used to eat James's agency weeks; now it's one prompt.
And if you have zero case studies? Don't sell. Build one funnel free for the best local business you can find, in exchange for a testimonial with numbers in it. Your cost is tokens and a weekend. Then you charge everyone after.
FAQ
Do I need to be a developer to sell AI funnels to local businesses?
No. You need a coding agent and taste. The build part (landing pages, follow-up sequences, research agents) is the easiest it has ever been. The scarce skill is direct response: writing an offer, matching the ad to the page, knowing why a $79 tripwire exists. Read Chet Holmes, Jay Abraham, and Dan Kennedy before you read another AI thread.
How much do local businesses actually spend on marketing?
More than most indie hackers think. Per the episode, a $5M plumbing company typically spends $150K to $250K a year, and single leads cost $129 in competitive trades. A $2,000 a month retainer is a rounding error next to a funnel that books a few extra $1,700 jobs.
Why funnels instead of starting an AI automation agency?
Because the customer doesn't want AI, they want booked jobs. James tested an AI receptionist on his own diesel business and customers hung up on it. Funnels attack a leak the owner can see (paid traffic hitting a page that can't convert) and the result shows up in their call log within weeks, not months.
What niche should I pick first?
High ticket, urgent, over $2M a year in revenue, and close to home. Plumbing and HVAC work but they're getting crowded by PE money. James's list of underpriced niches: hood cleaning, septic, tree services, gutter guards, sport court installs, backup generators. One trade, one city, six months.
This episode is the clearest "sell shovels to people who already struck gold" playbook I've heard this year.
And it's the same pattern I see every week on the podcast: the founders who win pick an unsexy market with money, solve one visible leak, and stay six months past the point where everyone else quits. I interview bootstrapped founders doing exactly that, with the real numbers.
