🎯 Learning Objectives
- Build a personal prompt framework for your domain
- Create industry-specific prompt playbooks
- Develop and maintain reusable template libraries
- Teach prompt engineering effectively to colleagues
- Measure and communicate the ROI of prompt engineering
- Prepare for the next generation: agents, tools, and automation
1. Building Your Personal Prompt Framework
MY PRE-FLIGHT CHECKLIST
Before running any prompt:
□ Did I assign a role?
□ Is the task crystal clear (verb + subject + purpose)?
□ Did I define the audience?
□ Are there length/format constraints?
□ Are there any conflicting instructions?
□ Have I specified what uncertainty should look like?
Framework Components
- Default role templates — the 5–8 roles most relevant to your work, pre-written
- Quality checklist — the 5 questions you ask before running any prompt
- Failure log — a record of prompts that failed and why
- Format library — output formats you use repeatedly, ready to paste
- Critique templates — standard critique prompts for your common output types
2. Industry-Specific Prompt Playbooks
PLAYBOOK: Pharmaceutical QA
─────────────────────────────────────────
Use Case 1: SOP Compliance Review
Prompt: [Full prompt] Score: 4.3/5.0
Review Required: Yes — QA specialist
Tested On: FDA 21 CFR Part 11, EU Annex 11
Use Case 2: Deviation Report Drafting
Format: ICH Q10 structure Score: 4.1/5.0
Review Required: Yes — QA manager sign-off
Use Case 3: Training Material Creation
...
─────────────────────────────────────────
3. Teaching Prompt Engineering to Teams
The Teaching Sequence
- Show the gap — Run a weak prompt on a task everyone recognises. Show the output. Then run the engineered version. The quality gap speaks for itself.
- Name the components — Identify which elements (role, task, context, constraints, format) were added and why.
- Practice with real work — Give each person a task from their actual workload. Have them write and improve a prompt live.
- Build the library — Collect the best prompts from the session. Version and share them immediately.
- Feedback loop — Create a channel for people to share prompts that worked, failed, and were improved.
4. Measuring ROI of Prompt Engineering
| Metric | Before | After | How to Measure |
|---|---|---|---|
| Time to first draft | X hours | Y hours | Time-track a sample |
| Revision cycles | X rounds | Y rounds | Count revisions on comparable tasks |
| Output quality score | X/5 | Y/5 | Blind expert rating of before/after |
| Hallucination rate | X per 10 | Y per 10 | Fact-check a sample of outputs |
💡 The Business Case
The most persuasive ROI argument is not time saved — it is risk reduced. One regulatory gap caught by a compliance review prompt can justify an entire programme investment.
5. The Future: Agents, Tools & Automation
- AI Agents — AI systems that plan and execute multi-step tasks autonomously. Prompt engineering defines the goals, constraints, and guardrails.
- Tool Use — AI systems that call external tools (search, databases, code execution). Prompts specify when and how to use each tool.
- Multi-agent systems — Networks of specialised AI agents working together. Prompt engineering defines the protocols between agents.
- RAG — AI that pulls from your organisation's knowledge base. Prompt engineering defines what to retrieve and how to synthesise it.
💡 The Enduring Principle
Regardless of how AI technology evolves: clarity of intent, precision of instruction, and human oversight of outcomes remain the permanent foundation of responsible AI use at {HP}.
✏️ Module 12 Exercise
Build your personal prompt framework: (1) your 5 most common work tasks that could benefit from AI, (2) a pre-written role template for each, (3) your quality checklist, (4) your first critique prompt. Share it with one colleague and improve it based on their feedback.
🔑 Key Takeaways
- A personal framework turns individual learning into a repeatable system
- Playbooks scale prompt engineering expertise across the organisation
- Teaching is most effective through live before/after demonstrations, not abstract principles
- ROI is most persuasively measured as risk reduced, not just time saved
- Prompt engineering is the foundational skill for the agentic AI era
- Clarity of intent, precision of instruction, and human oversight are permanent principles