Introduction & mindset

Realistic expectations and how agents really make money.

Goal: calibrate your expectations so you end up with success rather than disappointment. This is the
most important module. Don't skip it.

The big promise — and the honest truth

The dream: "I build an AI agent, turn it on, and it earns money while I sleep."

The reality, in one sentence:

**An AI agent doesn't earn money out of thin air. It automates and accelerates a real
business that delivers real value to real customers.**

That sounds less exciting, but it's exactly why this course actually works. We're not building a money machine; we're building a tireless employee that takes over the dull, repetitive work so your business can do more with fewer hours.

What an agent is genuinely good at today

• Producing content at scale (articles, product descriptions, social posts, emails).
• Answering customer questions (support, FAQ, lead qualification).
• Processing data (research, summaries, reports, spreadsheets).
• Executing routine tasks via tools (processing orders, scheduling appointments, following up).
• Working 24/7, in parallel, without breaks.

What an agent can't (yet) do

• Invent a business that has no market and conjure up customers anyway.
• Operate fully without oversight in a way that's legally and financially sound.
• Guarantee it will never make a mistake. (It will make mistakes. That's why guardrails exist — module 10.)
• Build trust, relationships, and reputation the way a human does (though it does help with all of these).

Whoever keeps these two lists separate will build something that generates income. Whoever mixes them up will burn through API budget and end up frustrated.

How agents actually make money: three levers

Money always comes from one of these three:

1. Saving time (reducing costs). You or your team do something in 5 hours; the agent does it in

5 minutes. You sell or reinvest that freed-up time.

1. Enabling scale (increasing revenue). You could handle 10 clients per month; with agents, 100 —

without hiring 10× more staff.

1. Offering a new service (new product). Something that wasn't profitable to do manually becomes

a sellable product with automation.

Every successful "AI agent business" falls under at least one of these levers. In module 02 you'll choose which one is yours.

Three levers that turn agent output into money
───────────────────────────────────────────────
  ┌──────────────────┐   ┌──────────────────┐   ┌──────────────────┐
  │  SAVE TIME       │   │  ENABLE SCALE    │   │  NEW SERVICE     │
  │                  │   │                  │   │                  │
  │  Agent does in   │   │  Serve 10x more  │   │  Automate what   │
  │  5 min what took │   │  clients without │   │  was too slow /  │
  │  you 5 hours     │   │  10x more staff  │   │  costly by hand  │
  └────────┬─────────┘   └────────┬─────────┘   └────────┬─────────┘
           │                      │                       │
           └──────────────────────┴───────────────────────┘
                                  │
                           ┌──────▼──────┐
                           │   REVENUE   │
                           └─────────────┘

The right mental models

1. Human-in-the-loop is not a weakness — it's a feature

The most powerful setup is not "100% autonomous," but "95% autonomous with human checkpoints at the right moments." The agent does all the work; you only approve the risky steps (spending money, publishing something, making a promise to a customer). This:

• prevents costly mistakes,
• keeps you legally safe,
• and builds confidence until you feel comfortable letting go of more.

💡 In Claude.ai: You can use a Claude.ai Project to store your agent's instructions and review its reasoning before any irreversible action. This is a great way to practice the "human checkpoint" pattern without writing a single line of code.

2. Start narrow, not broad

Don't build an "AI agent that can do everything." Build an agent that does one task excellently for one type of customer. An agent that writes perfect product descriptions for online stores beats an "agent that does marketing" by a wide margin.

3. Distribution beats technology

Building the agent is the easy part (you'll learn that here in a week). Finding customers is the hard part. Set aside time for distribution from day one: where are your customers, how do you reach them, why would they choose you? Modules 08 and 12 cover this.

4. Costs are real — do the math

Every agent action costs API tokens, which means money. An agent running unchecked in a loop can burn hundreds of dollars overnight. In modules 03 and 10 you'll learn how to control this and how to make sure your price always stays above your costs.

Realistic numbers (no get-rich-quick promises)

A few honest benchmarks so you know what's achievable:

• First working agent: an evening to a weekend.
• First paying customer: weeks to months, depending on your distribution.
• API cost per task: often cents to a few dollars. Your selling price needs to be well above

that (10×–100× is healthy for services).

• Time until "runs mostly on its own": months of tuning. Agents need maintenance,

just like employees do.

Whoever treats this like a serious business can build a serious income. Whoever expects a get-rich-quick button will quit after week two.

Your assignment for this module

Write — literally, in a notes file — your answers to these three questions:

1. Which lever (save time / enable scale / new product) do I want to use?
2. For whom do I want to create value? (as specific as possible: "furniture e-commerce stores,"

not "businesses")

1. What am I willing to invest in time and API budget over the next 30 days?

Save this. At the end of module 02 you'll make it concrete.