Module 3: AI-Assisted Instruction Writing
This practical module teaches how to leverage AI assistants like Manus and ChatGPT to efficiently create high-quality agent instructions. Rather than writing prompts from scratch, this approach uses AI to transform client specifications and templates into polished, effective instructions.
3.1 The AI-Assisted Approach to Prompt Engineering
Why Use AI Assistants for Instruction Writing? Creating effective agent instructions requires balancing multiple considerations: client requirements, agent capabilities, business context, and technical constraints. AI assistants excel at synthesizing these complex requirements into coherent, well-structured prompts.
Benefits of AI-Assisted Instruction Writing:
Faster development cycles from client specifications to deployed agents
Consistent quality and structure across different projects
Reduced human error in translating requirements to instructions
Easy iteration and refinement based on feedback
Standardized formatting that integrates seamlessly with raiaAI
The raiaLabs Workflow:
Gather client specifications and project requirements
Select appropriate instruction template from raia training materials
Use Manus or ChatGPT to generate initial instructions
Review and refine the generated instructions
Copy and paste the markdown-formatted instructions into raiaAI Launch Pad
Test and iterate based on agent performance
3.2 Working with Client Specifications
Typical Client Specification Components: Client specifications usually include several key elements that need to be translated into agent instructions:
Business Context and Objectives
Company background and industry
Target audience and customer personas
Primary business goals for the agent
Success metrics and KPIs
Functional Requirements
Specific tasks the agent should perform
Types of queries it should handle
Integration requirements with existing systems
Escalation procedures and limitations
Brand and Communication Guidelines
Tone of voice and communication style
Brand personality and values
Specific terminology or language preferences
Compliance and regulatory requirements
Technical Specifications
Deployment channels (chat, email, SMS, etc.)
Integration points and data sources
Performance requirements and constraints
Security and privacy considerations
3.3 Using raia Training Templates
Starting with Proven Templates The raia training materials include tested instruction templates for each agent type. These templates provide a solid foundation that can be customized for specific client needs.
Template Categories:
Support Agent Templates: For customer service and technical support scenarios
Sales Agent Templates: For lead qualification and sales conversion activities
Analyst Agent Templates: For data processing and workflow automation
Template Customization Process:
Select the template that best matches the agent type and use case
Identify sections that need client-specific customization
Prepare client specification details for each customizable section
Use AI assistants to merge template structure with client requirements
3.4 Prompting Manus for Instruction Generation
Effective Prompting Strategy for Manus: When using Manus to generate agent instructions, structure your prompt to provide comprehensive context and clear expectations.
Template Prompt for Manus:
I need you to create comprehensive AI agent instructions for a [AGENT TYPE] agent. Please generate the instructions in markdown format that I can copy and paste directly into raiaAI Launch Pad.
CLIENT CONTEXT:
- Company: [Company Name and Industry]
- Business Objective: [Primary goal for the agent]
- Target Audience: [Customer demographics and characteristics]
FUNCTIONAL REQUIREMENTS:
- Primary Tasks: [List of main functions]
- Query Types: [Types of questions/requests to handle]
- Escalation Rules: [When to escalate to humans]
- Integration Needs: [Systems to connect with]
COMMUNICATION GUIDELINES:
- Tone: [Professional, friendly, technical, etc.]
- Brand Voice: [Specific personality traits]
- Compliance: [Any regulatory requirements]
TEMPLATE BASE:
[Paste the relevant raia training template here]
Please create complete instructions that include:
1. Agent role and purpose
2. Communication style and tone
3. Knowledge source routing
4. Escalation procedures
5. Response formatting guidelines
6. Any specific business rules or constraints
Format the output in clean markdown that's ready to paste into raiaAI.Iterative Refinement with Manus: After receiving the initial instructions, use follow-up prompts to refine specific sections:
"Please enhance the routing section to be more specific about when to use each knowledge source"
"Add more detail to the escalation procedures for technical issues"
"Adjust the tone to be more conversational while maintaining professionalism"
3.5 Using ChatGPT for Instruction Development
ChatGPT Prompting Best Practices: ChatGPT can be equally effective for instruction generation when prompted correctly. The key is providing sufficient context and structure.
Sample ChatGPT Prompt:
Act as an expert AI prompt engineer specializing in customer service agents. I need you to create comprehensive instructions for an AI agent that will handle customer support for [COMPANY TYPE].
Here are the specifications:
[Paste client specifications]
Please create the instructions in markdown format, including:
- Clear role definition
- Communication guidelines
- Knowledge base routing instructions
- Escalation procedures
- Response formatting rules
The output should be ready to copy and paste into an AI agent platform. Make sure to include specific examples where helpful and maintain a professional but approachable tone throughout.Advantages of Each Platform:
Manus: Better integration with raiaAI ecosystem, understands platform-specific requirements
ChatGPT: Excellent for creative problem-solving and handling unique client requirements
3.6 Markdown Formatting for raiaAI Integration
Why Markdown Format? raiaAI Launch Pad accepts instructions in markdown format, which provides clean formatting and easy readability while maintaining structure.
Essential Markdown Elements for Instructions:
# Agent Role and Purpose
Clear definition of what the agent does
## Communication Style
Guidelines for tone and approach
### Knowledge Sources and Routing
- source1.md: For handling [specific query types]
- source2.md: For handling [specific query types]
## Escalation Procedures
When and how to escalate to humans
## Response Guidelines
- Always cite sources
- Maintain professional tone
- Provide clear, actionable information
## Specific Business Rules
Any client-specific requirements or constraintsCopy-Paste Workflow:
Generate instructions using Manus or ChatGPT
Review the markdown formatting for completeness
Copy the entire markdown content
Paste directly into the raiaAI Launch Pad instruction field
Preview and test the agent with sample queries
Iterate based on performance and feedback
3.7 Quality Assurance and Testing
Pre-Deployment Checklist: Before finalizing AI-generated instructions, verify:
All client requirements are addressed
Routing instructions reference actual knowledge base files
Escalation procedures are clearly defined
Communication style matches brand guidelines
Technical integrations are properly specified
Compliance requirements are included
Testing Generated Instructions:
Use sample queries that represent typical customer interactions
Test edge cases and unusual requests
Verify that routing works as intended
Check that escalation triggers function properly
Ensure responses maintain consistent tone and quality
Iteration and Improvement: Based on testing results, return to your AI assistant with specific refinement requests:
"The agent is too formal for our brand - please adjust the tone"
"Add more specific routing for billing-related questions"
"Include examples of how to handle angry customers"
This AI-assisted approach to instruction writing dramatically reduces development time while maintaining high quality and consistency across all agent deployments. The key is learning to effectively communicate requirements to AI assistants and then refining their output to meet specific client needs.
Last updated