Meet your new team member

Imagine you just hired someone who:

• Smart — has read practically every book, article, and paper ever published. Can speak on any topic.
• Fast — delivers in seconds what takes you 30 minutes
• Clueless about you — knows nothing about your role, your company, your preferences, or your current situation
• Occasionally wrong — can sound extremely confident while saying something completely made up (we'll cover how to catch this in Day 15)

That gap between "knows everything" and "knows nothing about you" — that's called context. And it's the single biggest lever you have. The more context you give, the better the output.

Context changes everything

Same AI, same question, completely different results:

Prompt frameworks that work

There are several proven structures for giving AI good context. Pick whichever feels natural — they all work. The point is: stop winging it.

Your task: Delegate 5 real tasks

Open your actual to-do list, inbox, or calendar. Pick 5 tasks. Click any card below to see the full prompt template you can copy-paste:

How to evaluate the output

After each task, ask yourself:

Why this matters

Professionals drown in documents. Reports, contracts, research papers, policy docs. Most get skimmed. Many don't get read at all.

AI changes this. Upload a 50-page PDF and get a structured summary in under 60 seconds. Not a vague overview — actual key points, decisions, and action items pulled out for you.

This is the single highest-ROI skill in this course. You'll use it every week for the rest of your career.

Step 1: Pick your document

Choose something real from your work. Not a test file — something you actually need to process.

Step 2: Upload and extract

Upload the file to your AI tool (drag & drop or attachment button). Then use one of these prompts:

Step 3: Go deeper

After the first summary, ask follow-up questions. This is where it gets powerful:

Why this matters

You think faster than you type. Way faster.

When you sit in front of a blank document, you edit while you write. You second-guess every sentence. A 30-minute task takes 2 hours because you're writing and thinking at the same time.

Voice flips this. You talk freely — messy, unstructured, stream of consciousness. Then AI organizes it. Two separate steps instead of one painful one.

Step 1: Choose your method

Step 2: Pick your brain dump topic

Choose something real you need to think through:

Step 3: Talk for 2-5 minutes

Step 4: Let AI structure it

Paste your raw dictation and use one of these:

Why this matters

Meetings generate information. But information without structure is noise.

Think about your last meeting. There were probably decisions made (not documented), tasks assigned (not tracked), deadlines mentioned (not written down), and ideas thrown around (never captured).

AI turns raw into structured. In seconds.

Step 1: Find your messiest notes

Grab one of these:

Step 2: Paste and extract

The transformation

Why this matters

You've been using AI for a week now. Some answers were great. Some were garbage. The variable wasn't the AI — it was you.

The difference between a useless answer and a brilliant one is almost always the prompt. Same model, same day, same topic — but one prompt gives you a generic blog post and the other gives you a strategy document your boss would actually use.

Today you learn why, and you practice the fix.

Step 1: Pick a real prompt

Choose something you've already asked AI this week. Something where the answer was "meh." Maybe it was:

• A draft email that sounded robotic
• A summary that missed the point
• Advice that was too generic to use

Copy that original prompt. That's your Version 1.

Step 2: The Prompt Improvement Formula

Each revision adds one layer. Follow this progression:

Step 3: Run all 5 versions

Paste each version into your AI tool, one at a time. Save all 5 outputs side by side. Look at the difference between Version 1 and Version 5. That delta? That's the skill.

The Prompt Improvement Formula (save this)

Why this matters

Every day this week, you've been typing context into AI from scratch. Your role, your company, your preferences — over and over.

That's like training a new intern every single morning.

Custom GPTs (OpenAI) and Gems (Google) solve this. They let you bake your instructions, context, and format preferences into a reusable AI program. You set it up once, then every time you use it, the AI already knows who you are, what you want, and how you want it.

This is where AI stops being a tool and starts being a system.

Pick your recurring task

Choose something you do at least once a week:

Build it: Custom GPT / Gem / Claude Project

Use the "Configure" tab. The key part is the instructions:

Why this matters

Most people use AI like a vending machine. Put in a prompt, get an output, walk away.

The good stuff happens in Round 2.

When you give AI feedback — specific, honest, direct feedback — it adjusts. And the second version is almost always better than the first. The third is better than the second. This is the feedback loop, and it's the difference between people who "use AI" and people who get real results from it.

Step 1: Open yesterday's Custom GPT / Gem / Project

Run the same task you tested yesterday. Get a fresh output. This is your Round 0 — the baseline.

Step 2: Give feedback (3 rounds)

Feedback phrases that actually work

Why this matters

Presentations eat time. Not because the thinking is hard, but because the format is unforgiving. You need a clear structure, concise text, a narrative arc, and visual consistency.

Most people start by opening PowerPoint and staring at a white rectangle. That's the worst way to start.

The better way: write a paragraph about what you want to say. Let AI handle the structure. You handle the editing.

Step 1: Write your brief

One paragraph. Describe what the presentation is about — who it's for, what's the main message, what do you want the audience to do after.

Step 2: Get the outline

Step 3: Fill at least 3 slides with real content

Step 4: Get design direction

Why this matters

Data is everywhere. In your CRM. In your project tracker. In that Google Sheet someone shared three weeks ago.

But data doesn't convince anyone. Narratives do.

"Revenue grew 14%" is a number. "Revenue grew 14% — driven entirely by one product line that didn't exist 6 months ago" is a story. The second one gets budget. The first one gets a nod.

Step 1: Find your data

Grab a real spreadsheet or CSV — anything with rows and columns:

Step 2: Upload and get the overview

Step 3: Find the outliers

Step 4: Write the narrative

The "So What?" test

After you have your narrative, read it and ask: "If my manager read this, would they know what to do next?"

If the answer is no, you have data analysis. Not a story.

Why this matters

Google gives you links. You click, scan, click again, open six tabs, lose track, and piece together an answer from fragments.

Perplexity gives you the answer. With numbered citations. In one page. No clicking required.

Step 1: Pick 5 things you'd normally Google

Real searches from your work or life this week:

Step 2: Run them through Perplexity

Go to perplexity.ai (free tier works). Try these prompt patterns:

Step 3: Compare with Google

For at least 2 searches, run the same query on Google. Note how many clicks Google took, whether Perplexity caught sources Google buried, and which answer was more actionable.

Step 4: Try Focus modes

When Perplexity wins vs. Google

Perplexity wins when you need a synthesized answer with citations, need to combine multiple sources, or want follow-up questions without starting over.

Google wins when you need a specific website, images, maps, or local businesses.

Why this matters

Quick chat answers are great for quick questions. But some questions aren't quick. "Should we expand into the Brazilian market?" "What's the best CRM for a 15-person sales team?"

Deep Research mode exists in ChatGPT, Gemini, and Perplexity. It takes 2-5 minutes. It searches dozens of sources, reads them, synthesizes the information, and produces a structured multi-page report. With citations.

The difference between chat and Deep Research is the difference between asking a friend and hiring an analyst.

Step 1: Pick your question

Good Deep Research questions are complex, specific to you, and not easily Googleable.

Step 2: Run it

ChatGPT: Click the Deep Research button. Gemini: Select "Deep Research" from model options. Perplexity: Use "Pro Search" or "Deep Research" mode.

Let it work. It takes 2-5 minutes. Don't interrupt it.

Step 3: Evaluate the report

Quick vs. Deep: A side-by-side

Why this matters

You have meetings tomorrow. Probably 3-5 of them. You know the title and the time. That's it.

You don't remember what was discussed last time. You haven't reviewed the documents. You don't have a clear goal for the meeting. You'll walk in, listen, react, and walk out wondering what you were supposed to push for.

What if AI prepped every meeting for you in 5 minutes flat?

Step 1: Choose your approach

Step 2: Generate meeting prep briefs

Step 3: Add context for high-stakes meetings

Why this matters

You have a pile of documents you should have read by now. Company policies. Research reports. Competitor analyses. Strategy decks from last quarter.

NotebookLM (from Google, free) changes the equation. Upload your documents. It reads all of them, understands how they connect, and lets you ask questions across multiple documents at once, generate study guides, and create Audio Overviews — podcast-style conversations about your documents.

Step 1: Gather 3+ documents

Pick documents that are related to each other. The magic happens when NotebookLM connects information across sources.

Step 2: Upload to NotebookLM

Go to notebooklm.google.com. Create a new notebook. Add your sources. Wait 30-60 seconds.

Step 3: Generate an Audio Overview

Click "Audio Overview" → "Generate." Two AI hosts will have a natural conversation about your documents — highlighting key points and making connections. Listen while doing something else.

Step 4: Ask questions across your documents

Why this matters

AI doesn't know what it doesn't know. And it never says "I'm not sure." It delivers wrong answers with the same confidence as right ones. Same formatting. Same tone.

This is the single biggest risk of using AI. Not that it's useless — it's that it's useful 95% of the time, so you stop checking. And then it fabricates a statistic, invents a citation, or confidently misstates something.

Today you learn to catch it. On purpose.

Step 1: Try to make it hallucinate

Step 2: The "Trust but Verify" framework

Every AI answer falls into one of three zones:

Step 3: Master the verification prompts

Your 10-second verification habit

Use this on any AI output you plan to share:

Why this matters

You interact with physical data every day. Receipts, business cards, whiteboards, handwritten notes, screenshots of dashboards, scanned contracts. All of it is trapped — visible to your eyes, invisible to your tools.

AI with vision can read images, extract the data, and output it in whatever structure you need — table, JSON, CSV, bullet points. A receipt becomes an expense line. A whiteboard becomes an action list.

Today you stop being a human OCR machine.

Step 1: Capture something real

Grab your phone and photograph one of these:

Step 2: Upload and extract

Step 3: Get specific formats

When to use this vs. when not to

Use AI OCR for: Quick extractions where 95% accuracy is fine. Internal docs, personal receipts, meeting notes.

Don't rely on it for: Legal documents where every character matters. Financial records needing audit-level accuracy. Apply Day 15's "trust but verify" for critical fields.

Why this matters

Content is the currency of professional visibility. But most people don't have a content problem — they have a production problem. The ideas are there. The time, the design skills, and the energy to turn one idea into multiple formats? Not there.

AI collapses the production bottleneck. This isn't about replacing creatives. It's about removing the excuse of "I don't have time to post."

Step 1: Pick your idea

Think of one thing you know about your work that most people don't. One insight, one lesson, one opinion. Write it in one sentence. That's your seed.

Step 2: Generate an AI image

Step 3: Turn your idea into a LinkedIn carousel

Step 4: Repurpose into 3 formats

Why this matters

There's a tool you've wanted for years. A calculator for that thing you estimate manually. A dashboard for the three numbers you check every Monday. A quiz for onboarding new team members.

You never built it because building things requires developers, timelines, and budgets.

That wall is gone. Claude's Artifacts feature lets you describe what you want in plain English and get a working, interactive application — running in your browser, shareable via link. No code. No deployment.

Step 1: Pick your project

Choose something you'd actually use. Not a demo — a tool that solves a real annoyance.

Step 2: Describe it in plain English

Step 3: Iterate on it

The first version will be 80% right. Now refine it:

Step 4: Polish and share

What you can build vs. what you can't

CAN build: Calculators, quizzes, simple dashboards with charts, form-based tools, interactive checklists, data visualizations, simple games.

CAN'T build (yet): Anything needing a database, tools connecting to external APIs, multi-page apps with complex navigation.

Why this matters

Everything you've done in this sprint has followed the same pattern: you ask AI something, it answers, you copy the answer somewhere else. You're the middleman.

Today that changes. You connect AI directly to the tools you use — so it can take action, not just generate text.

This is the difference between a chatbot and an assistant. A chatbot answers questions. An assistant does things.

Step 1: Pick your connection

Choose ONE integration. Don't try to connect everything — one working connection beats five half-finished ones.

Step 2: Set up and test

Step 3: Automate one real workflow

Step 4: See AI take an action

Ask AI to DO something, not just tell you something:

The AI assistant spectrum

Why this matters

Twenty days ago you opened an AI tool and sent your first real prompts. Today you prove — to yourself — that you're not the same user you were on Day 1.

This isn't a test with right answers. It's a capstone challenge: take a real problem from your work and solve it end-to-end using the skills from this sprint. No hand-holding. No step-by-step instructions. Just you, your AI stack, and a problem worth solving.

The Capstone Challenge

Pick a real problem from your actual work — something that's been sitting on your list. Then solve it using at least 3 skills from this sprint.

Skill menu (pick at least 3)

Example capstone projects

How to approach it

Your AI Stack — The Reflection

Before you close this course, answer these questions. Write them down — this is your personal AI operating system.

The full arc

• Week 1: Fundamentals — AI went from "thing I should try" to "tool I use daily."
• Week 2: Context — Better input = better output. Prompting, custom instructions, feedback loops.
• Week 3: Power Tools — Research-grade AI. Search, deep analysis, document intelligence, fact-checking.
• Week 4: Assistant — Your personal AI stack. OCR, content, tools, automations, and connections.

What's next?

• Keep your daily habits. Skills decay fast if you stop. Use AI for at least one real task every workday.
• Go deeper on your top 3. Use them daily for 30 days and they'll become instinct.
• Follow the tools, not the hype. Every month try ONE new feature from your primary tool.
• Teach someone. The fastest way to solidify what you learned. Pick one skill and teach it to a colleague this week.
• Build your prompt library. Save every prompt that worked well. In 3 months you'll have a personal playbook.