Page: 'Product Weekly — May 4, 2026'\n\nAttendees: Sara (PM), Marcus (eng lead), Priya (design), Tom (marketing)\n\nUpdates:\n- Sara: Onboarding redesign shipped Tuesday. Activation rate +18% on first cohort.\n- Marcus: API rate-limit bug from last week is fixed in main. Hotfix going out tonight.\n- Priya: Mobile nav prototype ready for review. 
  Need feedback by Friday so we can hand off to eng.\n- Tom: Q2 launch campaign needs hero copy. Stuck on positioning — should we lead with 'fast' or 'reliable'?\n\nDecisions:\n- Pricing experiment: A/B test the $19/$29/$49 ladder vs current $15/$25/$45 starting May 11. Sara to wire up GrowthBook flag.\n- Mobile rollout: gated behind feature flag, 5% rollout    
  next Monday, full rollout May 18 if metrics hold.\n\nAction items:\n- Sara: ship pricing flag by EOW.\n- Marcus: write postmortem for the rate-limit bug — due Wed May 7.\n- Priya: schedule mobile prototype review for Thursday.\n- Tom: draft 3 positioning options, share by Friday for team vote.\n\nOpen:\n- Do we hire a second backend engineer this quarter or  wait for Q3? CFO wants headcount frozen, but we're at 80% capacity. Tabled to next week.\n- Should the pricing experiment include an annual discount tier? Sara to ask David.

Summary  
This chapter demonstrates how to craft precise system prompts to guide an AI model in processing raw text into a structured format—extracting a TLDR, action items with owners and deadlines, and open questions.  

General domain of usage  
Automated meeting note summarization and action-item tracking

Useful AI at work isn't about searching the internet – it's about working on **your stuff**. Your meeting notes, your specs, your brief. That's what a tool like `notion_page` unlocks.



## The Shift

A search tool fetches the world. A page-reader tool fetches **your context**:

| Tool | Question It Answers |
|---|---|
| `web_search` | What's true **out there**? |
| `notion_page` | What did **we** decide last Tuesday? |



## What This Lab Does

Claude reads a page of meeting notes and turns it into something usable: a **TLDR**, a list of **action items** with owners, and the **open questions** that still need a decision. That's the kind of post-meeting writeup a chief of staff would charge for – Claude produces it in seconds, but only if the system prompt tells it exactly what shape to deliver.



## Why "Give Me a Summary" Isn't Enough

Take a page of meeting notes and ask any AI for **"a summary."** You'll get a paragraph – maybe useful, maybe not. Now ask for **"TLDR in 3 bullets, then action items with owner and deadline, then open questions"** and you'll get something you can **act on**.

**When a tool feeds Claude raw content, the system prompt is what turns content into action.**



## Task
Send this prompt:

><b>Read my Product Weekly notes from May 4 and give me a TLDR, action items with owners and deadlines, and the open questions.

Then push it with follow-ups – the notes are already loaded, so Claude doesn't re-fetch:

><b>Just show me what Sara owns this week.

><b>What decisions are blocked on someone outside the team?

><b>Which action items have no deadline?



A hands-on, beginner-friendly course designed for non-technical business users to understand and leverage AI tools and concepts in their daily work. The course covers foundational AI concepts, effective communication with AI, practical use of AI-powered tools, and building simple AI assistants.

Introduction to artificial intelligence, large language models, and essential AI vocabulary for business users.

Learn how to communicate effectively with AI systems through prompt engineering and tone selection.

Explore how AI can be integrated with business tools for research, note-taking, and data analysis.

Learn the basics of agent personas and how to build and tune simple AI assistants for work.