Evaluation Exercise: Prompt Development and Testing
Objective:
This exercise is designed to evaluate your ability to craft, implement, and validate a detailed prompt for
generating a structured set of actions tailored to specific user scenarios. The focus is on ensuring logical
clarity, exhaustive coverage, inclusion of prerequisites, and testing the prompt with real-life examples. The
actions generated by the AI agent should align with the user’s style, preferences, and priorities while
leveraging generative AI capabilities for efficiency and creativity.
Exercise Tasks
1. Prompt Creation:
Develop a high-quality prompt for generating a structured set of actions. The prompt must:
● Guide the AI agent to produce logical, comprehensive, and adaptable tasks.
● Include all prerequisites and dependencies for each action.
● Emphasize the importance of not missing any actions or steps.
2. Real-Life Examples:
Test the prompt using the following real-world processes. Your goal is to validate the prompt’s
ability to generate actionable outputs:
● Customer Support Workflow: Organize and prioritize support tickets with varying
urgency and complexity.
● Project Management Task Prioritization: Create a prioritized task list with deadlines
and dependencies.
● Product Launch Planning: Develop an exhaustive plan including milestones, approvals,
and fallback strategies.
● Team Communication Analysis: Summarize internal communication, analyze
sentiment, and prioritize responses.
● Data Cleanup and Migration: Generate a step-by-step plan for data validation, cleanup,
and migration.
● Marketing Campaign Execution: Outline a detailed multi-channel marketing campaign
execution plan.
3. Action Output:
The prompt must produce a structured output in table format with the following columns:
● No: Sequential numbering of actions.
● Category: Grouping or type of the action.
● Objective: Purpose of the action.
● Action: Detailed description of the task.
● Dependencies: Required conditions or prior tasks.
● Prerequisites: Necessary setup steps or inputs (e.g., approvals, credentials, system
integrations).
● Priority: Ranking based on urgency or importance.
● Risks and Mitigation: Challenges and solutions.
● Frequency: Task repetition schedule (e.g., daily, weekly, one-time).
4. Iterative Refinement:
Include a method to iteratively improve the prompt based on the quality of the output. Address
any gaps or ambiguities to ensure completeness and accuracy.
Evaluation Criteria
You will be assessed based on the following criteria:
1. Prompt Quality:
● Is the prompt clear, specific, and actionable?
● Does it align with the objectives outlined above?
2. Comprehensiveness:
● Does the AI-generated output cover all actions, prerequisites, and dependencies?
● Are there any steps or tasks missing?
3. Relevance and Applicability:
● Are the outputs relevant to the real-life examples provided?
● Does the prompt work effectively across all scenarios?
4. Logical Structure:
● Is the output well-organized and logically sequenced?
● Are dependencies and prerequisites accurately captured?
5. Adaptability:
● Can the prompt handle varying levels of complexity and edge cases?
● Does it generate modular, scalable actions?
6. Creativity and Innovation:
● Are generative AI capabilities effectively utilized?
● Are the outputs empathetic to user needs and priorities?
Deliverables
1. Prompt:
Provide the full text of the prompt you created for the exercise.
2. Testing Results:
Share the outputs generated by the AI when applied to the six real-life examples. Present them in
the specified table format.
3. Refinement Process:
Include notes on how you refined the prompt based on initial outputs and identified gaps.
4. Insights and Recommendations:
Provide a brief summary of your observations, challenges encountered, and suggestions for
improving the exercise.
Important Notes
● Ensure that no actions, prerequisites, or dependencies are missed.
● Focus on generating outputs that are clear, complete, and logically sequenced.
● Pay attention to adaptability across different scenarios.
Good luck!