The AI Foundations Bootcamp | Part Two: Stop Re-Explaining Yourself to AI.

Most AI users never touch the two features that make the biggest difference. Projects let AI remember your context so you stop re-explaining yourself every conversation. Deep research lets AI reason for 10-15 minutes before responding, producing work that looks nothing like a basic chat. This workshop shows you how to set up both. You'll leave with a working project you can use immediately and the confidence to deploy deep research on real problems. These two features alone will change how you work with AI.

Click here to grab the Prompt templates and Inspirational Use Cases for Projects and Deep Research

[00:00] Recap: Why AI Outputs Disappoint

• Trained on everything = defaults to bland average
• Talk it out method: have AI interview you before drafting
• Two context docs to stop re-explaining: personal profile + company profile
• These docs level up quality, accuracy, and relevance instantly

[08:00] Projects & Deep Research: The Underrated Power Features

• Hype cycle reality: interview prompting gets you up the slope of enlightenment
• Three questions for struggling AI users: paid plan? Interview method? Projects?
• Proposal example: generic prompt vs. project-loaded = night and day quality
• Five-minute setup pays off across unlimited repetitive tasks

[14:00] The Context Engineering Balancing Act

• Cold start problem: no context = generic outputs
• Context rot: conversation too long = confused, forgetful, wrong
• Visual model: context quality degrades as conversation extends
• Projects solve this by persisting instructions outside the chat window

[21:00] Building Your Project Step by Step

• Internal tokens: writing style guides, brand docs, SOPs, winning examples
• File format tip: Markdown preserves hierarchy, beats .txt and proprietary formats
• Knowledge capacity rule: stay under 50% to preserve conversation room
• Chunk large files (SOPs, sales data) into smaller, relevant segments

[29:00] Writing Project Instructions with AI

• Use a separate chat to draft your system prompt
• Instructions define: output format, file references, interaction process
• Test for accuracy, tone, and repeatability—then iterate
• If struggling, chunk the task smaller or add more context

[38:00] Choosing the Right Projects to Build

• Criteria: repetitive, has good examples, you can evaluate quality fast
• Stay in your lane—don't automate skills you haven't developed taste for
• LinkedIn posting mistake: no expertise = can't judge if output is good
• Boring but important tasks = easiest team buy-in

[45:00] Making Deep Research a Habit

• Like a washing machine for information work—runs while you do other things
• Platform locations: Claude (Research button), ChatGPT (Deep Research), Gemini (Deep Research)
• Always choose the most advanced model—this is where paid plans pay off
• NotebookLM bonus: deep research across web AND your Google Drive (coming soon)
• Use AI to write your research prompt first for better, more specific outputs