Start an AI automation agency.
Of every play in this track, an automation agency is the fastest route to real cash — you build automations that save a business hours every week, then charge a setup fee plus a monthly retainer to run them. Real 2026 retainers run $1,000–$10,000/mo, the demand is under-served, and you don't need to code. This is the full playbook: what you actually sell, the five workflows that pay, which tool to use, exactly how to price it (with a calculator), and how to land client one.
01 What you're actually selling
Read the anchor lesson and you already know the trap: nobody buys "AI." A plumber does not want an "AI agent." He wants to stop losing $2,000 jobs because his phone rang while he was under a sink. You sell the outcome — saved hours, captured leads, removed busywork — and the AI is just how you deliver it.
The business model has two parts, and the second is where the money compounds:
- Setup fee — a one-time project price to build and launch the automation. This pays for your time and de-risks the client ("pay to get it live").
- Monthly retainer — a recurring fee to monitor, maintain, and improve it. Automations touch live workflows that drift, break, and change; you're selling continued ownership of something that already works.
One-off builds make you a freelancer on a treadmill — always hunting the next project. The retainer is what turns this into an agency: predictable monthly revenue that stacks. Ten clients at $2,000/mo is $20k/month that doesn't reset on the 1st. Land the build, then sell the ownership.
Three "automation ideas." Which is something a business will actually pay a retainer for?
02 The five workflows that actually pay
You don't need to invent anything. The same handful of builds sell over and over, across industries — because they map to money a business is already losing. Master these and you have a product line:
| Workflow | What it does | Who buys it |
|---|---|---|
| Speed-to-lead | New lead → instant personalized reply + booked call + CRM entry | Anyone who buys ads or gets web leads |
| Inbox triage | Sorts inbound email, drafts replies, flags what needs a human | Owner-operators drowning in email |
| Meeting → action | Call recording → summary + CRM update + follow-up tasks | Sales teams, consultants, agencies |
| Document extraction | Invoices/forms/PDFs → clean structured data in a sheet or system | Bookkeeping, ops, logistics, legal |
| Content repurposing | One long input → clips, posts, newsletter, show notes | Creators, coaches, marketing teams |
Notice they're all the same shape: a trigger, an AI step that does judgment, and an action in a tool the business already uses. That shape is the entire craft. The first one — speed-to-lead — is the one I'd build first every time, because the ROI is dead obvious: a lead answered in 2 minutes is 21× more likely to convert than one answered in 30. You're not selling automation; you're selling the leads they're currently letting rot.
03 Pick your tool
You'll build on a no-code/low-code automation platform wired to an LLM (OpenAI or Anthropic) for the "thinking" steps. Three platforms own the market — and the right one depends on you, not on which is "best."
| Tool | Best for | The trade |
|---|---|---|
| Zapier | Fastest to start; the most app integrations | Gets expensive fast at real task volume |
| Make | The agency sweet spot — visual, powerful, affordable at scale | Slightly steeper learning curve than Zapier |
| n8n | Cheapest at scale; self-hostable; developer-friendly | More technical; you manage more yourself |
Pick what's most true for you:
04 Price it (the part people get wrong)
The number one reason new agencies starve isn't bad tech — it's charging like a freelancer. You're not selling hours; you're selling a result that's worth thousands a month to the client. Price against their ROI, not your time. Here's the shape of a real 2026 deal — drag the inputs and see:
The move that makes the retainer easy to sell: land the build first, then convert. Charge a setup fee to get the workflow into production. Once it's live and visibly saving them money, the retainer isn't a promise about the future — it's continued ownership of something already paying for itself. That conversation closes itself.
05 Land client one (the real bottleneck)
Everything above is the easy 80%. This is the 80% that pays — and it's why most people who "learn AI automation" never make a dollar. The build is a weekend; the client is the business. Three rules:
- Niche down hard. "I automate businesses" is invisible. "I make sure roofing companies never miss a lead" is a referral magnet. One industry, one painful problem. You can expand later; you can't get traction broad.
- Lead with a free audit, not a pitch. Offer to map where a specific business is losing time or leads — a 20-minute loom or call. You'll spot three automations and have earned the right to quote one.
- Sell the leak, not the tech. "You're getting ~40 web leads a month and answering them in a day — that's roughly 8 jobs walking out the door. I'll get every one answered in 2 minutes." Numbers, their money, their words.
Pick a niche you could actually talk to:
Your move this week
Pick one niche and one workflow (speed-to-lead is the safe bet). Build it once for a fake business so you have a working demo, then offer a free audit to five real businesses in that niche. You're not selling yet — you're collecting the language and landing one pilot for a case study. The case study is what makes client two through ten easy.
What you can do now
- Sell outcomes (saved hours, captured leads), never "AI"
- Lead with the setup fee, then convert to the retainer that compounds
- Build from the five proven workflows — start with speed-to-lead
- Pick one tool (Make is the agency default) and get fluent instead of tool-hopping
- Price against the client's ROI, not your hours — use the calculator's middles as a floor
- Niche down, open with a free audit, and sell the leak in their own numbers