# AI Tools Guide

All prompts in this library work across ChatGPT, Claude, and Microsoft Copilot. Here are the practical differences to know.

---

## Quick Comparison

| | ChatGPT (GPT-4o) | Claude (Anthropic) | Microsoft Copilot |
|--|-----------------|-------------------|-------------------|
| **Best for** | Long-form synthesis, structured tables | Nuanced coaching language, nuanced analysis | M365 integration, Teams/SharePoint |
| **Context window** | Large | Very large | Varies by plan |
| **Tone quality** | Strong | Excellent for coaching/facilitation | Good |
| **Integration** | API, browser, mobile | API, browser, mobile | Embedded in M365 |
| **Table formatting** | Excellent | Excellent | Good |
| **Hallucination risk** | Low with good prompts | Low with good prompts | Low with good prompts |

---

## ChatGPT (GPT-4o)

**Best use cases in this library:**
- Long retro synthesis with many inputs
- Multi-sprint pattern analysis
- Generating multiple versions of Sprint Goals or PI Objectives
- Structured tables (dependencies, ROAM risks)

**Tips:**
- Use the system prompt feature (in Custom Instructions) to pre-set your Scrum context once
- GPT-4o handles large pastes of retro sticky notes very well
- Use "Code Interpreter / Data Analysis" mode if you're working with metrics

**Watch out for:**
- Overconfident output — always review PI Objectives for accuracy before sharing
- Verbose responses — add `"Be concise"` or `"Use bullet points only"` if needed

---

## Claude (Anthropic)

**Best use cases in this library:**
- Retrospective facilitation prompts (nuanced, empathetic tone)
- Coaching language for action items and improvement areas
- Stakeholder narratives and sprint review summaries
- Complex dependency analysis

**Tips:**
- Claude handles ambiguous team context well — it will ask clarifying questions if something is unclear
- Excellent at varying tone across audiences (team vs. leadership vs. Business Owners)
- Great for follow-up conversations: paste Claude's output back and ask for refinements

**Watch out for:**
- Responses can be longer than needed — specify your length requirement in the prompt
- May add caveats you don't need for internal documents — prompt with `"Write this as a direct summary, no caveats"`

---

## Microsoft Copilot (M365)

**Best use cases in this library:**
- Sprint review and retro summaries directly into Confluence/Word/SharePoint
- Teams post drafts for stakeholder updates
- Meeting summaries from Teams recordings (Copilot in Teams)
- PI Objective tables in Word or PowerPoint

**Tips:**
- Use Copilot in Teams to auto-summarise PI Planning or retro meetings — then use the prompts here to structure that summary
- Copilot in Word: paste the prompt + context directly into Word, select it, and use Copilot to rewrite/expand
- Copilot Pages: useful for drafting living team documents from prompt outputs

**Watch out for:**
- Quality varies by M365 plan (Business vs. Enterprise)
- Works best when you already have the context in your M365 ecosystem (notes, emails, calendar)

---

## General Tips for All Tools

1. **Provide more context than you think you need** — AI models perform much better with specific inputs
2. **Iterate, don't restart** — follow-up prompts within the same conversation preserve context
3. **Ask for alternatives** — `"Give me 3 different versions of this Sprint Goal"`
4. **Check accuracy** — AI can confidently produce plausible-but-wrong output, especially for team-specific details
5. **Use AI for drafts** — let AI do the first 80%, then apply your Scrum Master judgment for the last 20%