AI Prompts

Writing Prompts

See “Improve Prompt Results” section of Microsoft Azure AI Fundamentals, Generative AI - Microsoft Azure AI Fundamentals: Generative AI from Improve prompt results - Training | Microsoft Learn

Good prompts provide the AI with:

Goal (what you want), see tasks below
Context (people involved, why you need something), see job types below
Information sources to use (attachments, documents, links, meetings)
Expectation on what you want (tone, type of output, format, response length and details like sources, examples of results you want like few shot method)
Iterate on the previous prompt

Other tips:

Use positive instructions, tell model what to do instead of what not to do
Be specific and detailed and avoid “what does this do” where “this” is unclear or prompts with a large scope of possible results
Use terms likely to appear in the context or expected results

Example Task Template

In Markdown

## TASK: <Short Action Summary>

**ID:** task-2026-04-14-001
**Status:** Queued
**AI Role:** Analyst
**AI Model:** auto (automatically select best model)
**Output Type:** Written Analysis

---

### 1. Goal
Clearly state what you want accomplished.
- Desired outcome
- Definition of complete

---

### 2. Context
- History
- Audience / people involved
- Constraints (time, politics, sensitivities, costs)

---

### 3. Information Sources
- **Attachments:**
  - doc1.pdf
  - requirements.md
  - python_code.py

- **Links:**
  - [1]

- **Meetings / Notes:**
  - [2]

---

### 4. Expectations
- **Tone:**
- **Output format:**
- **Length:**
- **Structure:**

In Emacs Org Mode - <Short Action Summary>

,:PROPERTIES:
,:ID: task-2026-04-14-001
,:STATUS: draft
,:AI_ROLE: analyst
,:AI_MODEL: gpt-4.1
,:OUTPUT_TYPE: written_analysis
,:END:

**** 1. Goal
Clearly state what you want accomplished.
- Desired outcome
- Definition of task complete

**** 2. Context
- History
- Audience / people involved
- Constraints (time, politics, sensitivities, costs)

**** 3. Information Sources
- Attachments: doc1.pdf, requirements.md, python_code.py
- Links:
  - [1]
- Meetings / Notes:
  - [2]

**** 4. Expectations
- Tone:
- Output format:
- Length:
- Structure:

Example - Create Cloud Architecture Decision Summary

1. Goal

Produce a concise architecture decision summary suitable for senior leadership.

Success = clear recommendation with pros and cons.

2. Context

Audience: Director level
Decision impacts current fiscal budget
Prior discussion was fragmented

3. Information Sources

Architecture options doc (arch-options-v3.pdf)
Cost estimates spreadsheet
Meeting notes from 2026-04-14

4. Expectations

Tone: Neutral, professional
Format: 1 page memo
Structure:
- Background
- Options
- Recommendation
Length: ~400 words
Examples: AWS decision memos

Tasks and Jobs for Prompts

Example Tasks and Job Types for Prompts from Microsoft Copilot Prompts Gallery with tips, examples with my additions

Tasks

Catch up
Create
Ask
Understand
Learn
Schedule
Edit
Prepare
Analyse
Code
Find
Design
Execute
Ideate

Job Types

Department

Accessibility
Frontline Management
Executive
Human Resources
Marketing
Sales
Communications
Marketing
Operations
Finance
Project Management
Information Technology
Customer Service
Legal

Industry

Manufacturing
Retail
Sustainability
Financial Services
Energy
Consumer Goods
Mobility
Nonprofit
Government
Healthcare
Media and Entertainment
Education - Student / Faculty

Agentic Prompts

Write a file like in markdown with:

Problem / Goal
Explain agent’s role
Tasks to follow
Response expectations
Example results and/or resources for agent to refer to