How to Prompt Like a Pro: Master AI for Business Innovation

An O'Reilly Live Learning course teaching business professionals how to extract maximum value from AI tools like Microsoft 365 Copilot, ChatGPT, Claude, and Google Gemini.

Last updated: April 2026

Course Segments

#	Segment	Duration	Key Topics
1	Identity, Mindset & Context Foundations	50 min	Pilot/copilot, anchor trap, prompt smell, context disclosure, inference
2	Context Sculpting & Prompting Technique	50 min	Role-play, task decomposition, few-shot, chain-of-thought, meta-prompting
3	Workflow, Multimodal & Security	50 min	Custom instructions, prompt versioning, voice, cross-referencing, privacy
4	Agentic Orchestration & Resilience	50 min	LLM matching, subagents, checkpoints, MCP, Copilot Studio, breaking changes

Warner's Laws of Generative AI Prompting

You are the pilot; the AI is your co-pilot. You're responsible for its actions.
Always know who you're signed in as and who you're chatting with.
Beware the anchor trap — draft before you prompt.
Trust your gut — never hesitate to second-guess the AI.
Every AI chat has its own lifecycle; develop your "prompt smell."
The more you disclose in trust, the more the AI can help you.
Anything you leave out of your prompt will be inferred by the AI.
Surgically sculpt your context. Just because you can doesn't mean you should.
Role play like you're a director.
Don't swallow the elephant — break down complex tasks with the AI.
Show, don't tell — lead with examples.
Make the AI show its work.
Think meta: prompt about prompting and custom instructions.
Strike while the iron's hot.
If you need to remind the AI of something, add it to custom instructions.
Periodically refactor your custom instructions and memories.
Treat prompts as assets — version-control them.
Use your voice if using words is difficult.
Pick up a good book on technical writing.
Always have a trusted LLM to cross-reference responses.
Protect your LLM against abuse by integrating test prompts.
Protect privacy ruthlessly. Know your chat storage, licensing, and data retention.
Each LLM has its own personality. Match the tool to the task.
Orchestrate subagents as force multipliers. Use git worktrees for parallelism.
Checkpoint before consequence — autonomous does not mean unsupervised.
Expect breaking changes. Stay agile, adaptable, and be an eternal learner.

Context Engineering vs Prompt Engineering

Prompt Engineering	Context Engineering
Focus on what to say	Focus on what the model knows
One-off interactions	System-wide reliability
Phrasing and examples	Everything in the context window

Key Insight: Most AI failures aren't model failures—they're context failures.

Tools Covered

Primary Platforms

Microsoft 365 Copilot — Notebooks, Agents, enterprise integration
ChatGPT — Projects, Custom GPTs, DALL-E 3, Vision
Google Gemini — Gems, Imagen 3, 1M+ token context
Claude — Projects, long-form analysis, Claude Code

Agentic AI

Claude Code — Terminal-based autonomous coding with checkpoints
GitHub Copilot Coding Agent — Issue-to-PR cloud automation
M365 Copilot Studio — Enterprise multi-agent orchestration
Azure AI Foundry — Azure-hosted model deployment and orchestration

Prerequisites

Required

Internet connection
ChatGPT free account
Google account

Quick Start

CRAFT Framework

Context: [situation]
Role: You are a [role]
Action: [task]
Format: [output format]
Tone: [voice/style]

Example Prompt

Context: I'm preparing a quarterly business review for my VP.
Role: You are a senior business analyst.
Action: Analyze this sales data and identify the top 3 trends.
Format: Executive summary with bullet points, max 200 words.
Tone: Professional and data-driven.

Repository Structure

docs/                               # Reference guides + slide deck
images/                             # Cover art, social preview assets
segments/
├─ segment-1-identity-mindset-context/      # Laws 1-7, anchor trap, context foundations
├─ segment-2-context-sculpting-technique/   # Laws 8-14, few-shot, chain-of-thought
├─ segment-3-workflow-multimodal-security/  # Laws 15-22, versioning, privacy
└─ segment-4-agentic-orchestration/         # Laws 23-26, subagents, MCP demos
.github/                            # Issue templates, workflows, AI instructions
COURSE-PLAN-APRIL-2026.md          # April 2026 delivery plan

For instructors: See INSTRUCTOR-MANIFEST.md for delivery guide.
For agents: Review AGENTS.md and CLAUDE.md before editing lessons.

Resources

License & Contributing

Licensed under MIT. See CONTRIBUTING.md for guidelines and AGENTS.md for the agent-focused repository playbook.

Code of Conduct

Participation in this project is governed by the Code of Conduct.

Security

Found a vulnerability or risky prompt scenario? Follow the disclosure steps in SECURITY.md or email Tim directly at [email protected].

Repo Guides & Automation

AGENTS.md - Contributor playbook
CLAUDE.md - Copilot instructions
INSTRUCTOR-MANIFEST.md - Run-of-show notes
COURSE-PLAN-APRIL-2026.md - April 2026 delivery plan
markdownlint.json & Markdownlint workflow - spacing rules + one-click lint/autofix

Ready to prompt like a pro? Let's transform how you work with AI!

prompt-pro