Building Your AI Workflow

Meet Sarah's Tuesday — before and after AI

Before AI (10 hours):

After AI (6 hours):

Sarah didn't replace herself. She replaced the low-value parts of her day. The thinking, strategizing, and decision-making — the parts that actually require a human brain — now get her best hours instead of her leftovers.

That's what this module helps you build: your own version of Sarah's after.

Step 1: Audit your workweek

You can't optimize what you haven't mapped. Before building any AI workflows, you need to see exactly where your time goes.

The task audit method

For one full week (or even just one representative day), log every task you do. For each task, note:

1. What it is (in 5 words or fewer)
2. How long it takes
3. How often you do it (daily, weekly, monthly)
4. The type of work — use these categories:

Step 2: The Task-AI fit matrix

Now that you've mapped your day, it's time to sort each task into one of three categories. Not everything should involve AI. Some things absolutely should. And some should stay 100% human.

Three categories for every task

There Are No Dumb Questions

"What if I'm not sure whether a task is 'AI does it' or 'AI helps'?"

Start with "AI helps." Have AI do a first draft while you watch. If the output is 80% usable with minimal edits, move it to "AI does it." If you're rewriting most of it, it stays in "AI helps." If the AI output is useless, it moves to "you alone." Let experience decide, not theory.

"Won't my boss think I'm slacking if AI is doing my work?"

Your boss doesn't care how the report gets made — they care that it's accurate, on time, and insightful. If AI helps you deliver better work faster, that's a performance improvement, not slacking. The hours you save should go into higher-value work that your boss has been wanting from you. That's how you turn AI from "less work" into "better work."

Step 3: Build your playbook

A playbook is your personal collection of saved prompts, templates, and workflows. It's the difference between using AI sporadically ("let me try asking AI...") and using AI systematically ("let me grab my template for this").

What goes in your playbook

Building a saved prompt

A saved prompt uses the same 4-part formula you learned in Module 2 — but with Context designed as a fill-in-the-blank slot:

1. Role: Who should AI pretend to be?
2. Context: [PASTE YOUR INPUT HERE] — the part that changes each time
3. Task: What exactly should it do?
4. Format: What should the output look like?

Example — meeting notes prompt:

Role: You are a professional executive assistant summarizing
a meeting for busy stakeholders.

Task: I'll paste a meeting transcript. Extract:
- Key decisions made (bullet points)
- Action items with owner and deadline
- Open questions that need follow-up
- One-sentence meeting summary

Format:
## Meeting Summary
[one sentence]

## Decisions
- [decision 1]
- [decision 2]

## Action Items
| Action | Owner | Deadline |
|---|---|---|
| ... | ... | ... |

## Open Questions
- [question 1]
- [question 2]

Save this in a document, note app, or text expander. When a meeting ends, paste the transcript, paste the prompt, get your notes in 30 seconds.

Building a template

A template is a prompt with blanks:

Write a [TYPE] email to [RECIPIENT ROLE] about [TOPIC].
Tone: [professional / friendly / urgent].
Key points to include:
- [POINT 1]
- [POINT 2]
- [POINT 3]
Keep it under [NUMBER] words.

Building a workflow

A workflow chains multiple AI steps together:

Example — "New project research" workflow:

There Are No Dumb Questions

"Where should I store my playbook?"

Anywhere you can find it quickly. A Google Doc works. A Notion page works. A folder of text files works. Some people use text expander tools (like TextExpander or Espanso) so they can type a shortcut and the prompt appears. The best system is the one you'll actually use.

"How many prompts do I need?"

Start with three: one for your most common written task (emails, reports, etc.), one for meetings, and one for research. That covers 80% of most knowledge workers' AI usage. Add more as you discover new use cases — but don't try to build 20 prompts on day one.

Step 4: Measure the impact

If you can't measure it, you can't prove it. And if you can't prove it, your boss thinks you're just playing with chatbots.

What to measure

The one-week measurement challenge

Track these numbers for one normal week:

Weekly totals:

• Total time saved: ___ hours
• Tasks where quality improved: ___
• New things accomplished with saved time: ___

Presenting your results to your boss

When you inevitably want to share what you've learned (or justify your AI subscription), use this structure:

Before AI: I spent [X hours/week] on [task category].
After AI: It takes [Y hours/week] — a [Z%] reduction.
I've reinvested that time into [strategic work].
Result: [specific outcome — project completed faster,
higher client satisfaction, etc.]

Concrete is better than abstract. "I saved 4 hours per week" is good. "I saved 4 hours per week and used that time to complete the Q3 campaign plan two weeks early" is much better.

Your personal pit crew

Here's the final analogy that ties everything together. You're a race car driver. AI is your pit crew.

No pit crew ever won a race by itself. And no driver ever won without one. The combination is what wins.

Back to Sarah's Tuesday

When Sarah mapped her day using the task audit from Step 1, she found that 5 of her 9 tasks fell squarely into the "AI does it" category — email drafting, meeting notes, status reports, competitor research summaries, and the analytics write-up. Together those tasks consumed roughly 4.5 hours of her day. After building a three-prompt playbook and running it for two weeks, those same tasks took under 90 minutes total. The hours she recovered didn't disappear into more busywork — she deliberately protected them for strategic planning, the kind of thinking she previously squeezed into Friday afternoons when she had nothing left. Her boss noticed the Q3 campaign plan landing two weeks early; Sarah knew it was because AI had given her best hours back. The workflow didn't make her dispensable. It made the part of her that no AI can replicate — judgment, client relationships, creative strategy — the part that showed up first.

Sarah's Tuesday, revisited

Sarah didn't get a different job. She didn't work fewer hours. She redesigned how she worked — and the thing that changed wasn't the output, it was where her best thinking went.

The routine drafts, the meeting notes, the status updates: AI handled those. The strategy, the judgment, the client relationships: those got her best hours instead of her leftovers.

That's the whole point of this module. Not to automate everything — but to clear enough space that you can finally do the work that only you can do.

Key takeaways

• Audit before you automate. Map your actual workday before deciding what AI should touch. Most people overestimate how much AI should do and underestimate how much time they spend on automatable tasks.
• Every task falls into three categories: AI does it (you review), AI helps (you collaborate), or you alone (AI stays out). Let experience, not theory, sort your tasks.
• A playbook turns sporadic AI use into systematic AI use. Three saved prompts — for emails, meetings, and research — cover 80% of most professionals' AI needs.
• Measure the impact, then reinvest the savings. Time saved isn't the goal — doing better work with that time is. Track your numbers and redirect saved hours into strategic work your boss actually cares about.