Prompt Training — Lesson 2
Role, Context, and Task Design
Learning Objectives
- 1Design effective role assignments for different use cases.
- 2Select and structure relevant context without overwhelming the prompt.
- 3Write precise task instructions that produce actionable output.
Designing effective roles
The role you assign shapes how the AI approaches the task. A role includes expertise level, perspective, and sometimes personality. "Act as a skeptical venture capitalist evaluating a pitch" produces very different analysis than "Act as a supportive business mentor coaching a first-time founder."
Effective roles are specific. "Act as an expert" is too vague. "Act as a B2B SaaS marketing strategist with 15 years of experience in mid-market enterprise sales" gives the AI a specific lens through which to generate content. The role should match the expertise your task actually needs.
You can also use the role to set the output style. "Act as a concise technical writer who avoids jargon" produces different writing than "Act as an engaging storyteller who uses vivid analogies." Match the role to your audience needs.
Structuring context effectively
Context is the background information the AI needs to produce relevant output. Too little context produces generic results. Too much context dilutes the signal and can push important details out of the context window.
Effective context selection: Include information that directly affects the output — audience details, relevant facts, constraints, history, and source material. Exclude information that is interesting but does not change what the AI should produce.
When providing source material (documents, data, meeting notes), format it clearly. Use headings, separate sections, and label each source. Tell the AI what the source material is and how it should use it: "Below is the client feedback from our Q3 survey. Use these specific comments to support your analysis."
Writing precise tasks
A precise task includes an action verb, a deliverable, and success criteria. "Analyze our pricing page and suggest three specific improvements to increase trial signups, with rationale for each suggestion" is precise. "Look at our pricing page and give feedback" is not.
When the task has multiple parts, number them. This ensures the AI addresses each component and makes the output easier to review. "1. Summarize the key findings. 2. Identify the three most actionable recommendations. 3. List risks and mitigation strategies for each recommendation."
For complex tasks, break them into sequential prompts rather than trying to accomplish everything in one prompt. First ask for analysis, review it, then ask for recommendations based on the analysis. Multi-step conversations often produce better results than single comprehensive prompts.
Case Study
The role that changed everything
Situation
A nonprofit director asked AI to "review our donation page." The output was generic website feedback. She then asked: "Act as a conversion optimization specialist who works with nonprofits. Review our donation page copy (pasted below). Our average donor is 45-65, gives $50-200 annually, and cares about local impact. Identify the three biggest barriers to completing a donation and suggest specific copy changes for each."
Analysis
The specific role (conversion specialist for nonprofits), the audience details (age, giving range, values), and the precise task (three barriers with specific copy changes) transformed a generic review into actionable, relevant recommendations.
Takeaway
The role does not just change the tone — it changes the entire analytical framework. A conversion specialist notices different things than a general reviewer.
Reflection Questions
- 1. For a task you frequently delegate, what role, context, and task description would you include in a prompt?
- 2. Have you tried using different roles for the same task? How did the outputs differ?
Key Takeaways
- ✓Specific roles shape the AI analytical framework, not just the tone.
- ✓Include context that affects the output; exclude context that does not change what the AI should produce.
- ✓Precise tasks include action verbs, deliverables, and success criteria.
- ✓Break complex tasks into sequential prompts for better results.