Duration: 3–4 days | 6 lessons

Your Workflow Has a Pulse. Now It Needs a Brain.

You've completed Module 2. Your Morning Brief workflow is running. Every morning, it wakes up, fetches weather data, grabs the latest news headlines, and lands them in your inbox. It works.

But here's the thing: it's sending you data, not information. You're getting a list of temperatures, conditions, and headlines. You still have to mentally process it all. "Okay, 62 degrees and rain, so I need an umbrella. That headline about tech stocks is interesting. That one about celebrity gossip I can skip." You're doing the hard work.

In Module 3, you add a brain to your workflows. An LLM step that reads the raw data and tells you what actually matters. "Good morning. It's 62°F and rainy — bring an umbrella and grab a raincoat if you're heading out after 3pm. Three big stories today: tech earnings are up, markets are steady, and there's news from your industry about new regulation."

That's not just data. That's insight. And your workflow is doing the thinking, not you.

What You'll Learn

✅ How to add an AI step to a workflow — The mechanics: connecting an LLM API, configuring a prompt, and mapping data through it like any other module.

✅ How to prompt for automation, not conversation — Writing prompts that work unattended, handle edge cases, and produce reliable output every single run.

✅ How to get structured data out of AI — Using JSON and response formatting so downstream steps can reliably parse and use the AI's output.

✅ When and how to chain multiple AI steps — Combining AI calls in sequence to handle complex workflows without overcomplicating or breaking your budget.

✅ How to defend against hallucinations at scale — Because in a workflow, an AI error can run 100 times before you notice it.

✅ How to manage costs and avoid rate limit surprises — Because AI isn't free, and at scale you need to understand the bill.

Module Lessons

1. Lesson 1: AI as a Module — Connecting the Brain — Adding an LLM call as a workflow step. API keys, platform configuration, testing with real data.

2. Lesson 2: Prompting for Automation — Writing for the Machine, Not the Human — The difference between conversational prompts and automation prompts. Building robust system prompts. Iteration and edge case handling.

3. Lesson 3: Structured Output — Getting Data You Can Actually Use — Why JSON matters. How to request and parse structured output. Handling failures gracefully.

4. Lesson 4: Chaining AI Calls — When One Step Isn't Enough — Multi-stage AI processing. When chaining makes sense and when it's overkill. Cost considerations.

5. Lesson 5: The Hallucination Problem in Pipelines — When AI Gets It Wrong at Scale — Why unattended hallucinations are risky. Practical defenses: constrained outputs, verification, logging.

6. Lesson 6: Cost and Rate Limits — The Bill at the End of the Month — Understanding pricing and rate limits. Practical cost management. Monitoring and budgeting.

Module Deliverable

Two fully functional, documented AI-enhanced workflows:

1. Upgraded Morning Brief — Your existing Morning Brief workflow, enhanced with an AI step that transforms raw weather/news data into a 2-3 sentence plain-English briefing that highlights what's actually worth your attention.

2. Inbox Triage — A new workflow that runs every time an email arrives. It classifies the email (action-required / FYI / marketing / spam), summarizes it in one sentence, and writes the results to a Google Sheet. You'll have a living, searchable log of your inbox intelligence.

By the end of this module, you'll have workflows that don't just collect data — they think about it.