Agent Configuration Guide

Master the art of configuring AI agents with clear understanding of purpose vs system prompt and proven frameworks for optimal performance.

Overview

AI agents in DealGate are powerful tools that can be customized for specific business needs. Understanding how to properly configure your agents is crucial for getting the best results from your AI-powered workflows.

Purpose vs System Prompt: Understanding the Difference

Purpose (User-Facing Description)

The Purpose is a clear, human-readable description of what your agent helps with. This is what users see when they interact with your agent and helps them understand the agent's role and capabilities.

Purpose Characteristics:

  • User-facing: Shown in agent lists and chat interfaces
  • Clear and concise: Easy to understand at a glance
  • Business-focused: Describes the business value or use case
  • Marketing-friendly: Helps users choose the right agent

Purpose Examples:

  • "Find and qualify B2B software leads in London"
  • "Research competitors and market trends for SaaS companies"
  • "Analyze sales data and create performance reports"
  • "Generate follow-up emails for warm prospects"

System Prompt (AI Instructions)

The System Prompt contains technical instructions that tell the AI how to behave, what tools to use, and how to respond. This shapes the agent's personality, capabilities, and decision-making process.

System Prompt Characteristics:

  • AI-focused: Direct instructions to the AI model
  • Technical: Controls behavior, tools, and response patterns
  • Detailed: Specific guidelines for handling different scenarios
  • Personality-shaping: Defines tone, style, and approach

System Prompt Examples:

You are a professional B2B lead generation specialist working with software companies. 

Guidelines:
- Always validate email addresses before suggesting them
- Prioritize companies with strong online presence and clear contact information
- Use a consultative, professional tone in all interactions
- When presenting leads, include company size, funding status, and key decision makers
- Focus on companies with 10-500 employees in target markets
- Verify company information through multiple sources when possible

Tool Usage:
- Use email validation tools for all contact information
- Leverage business search tools to find comprehensive company data
- Store important findings in memory for future conversations
- Export verified leads in structured formats

Framework for Writing Effective Purposes

Purpose Framework: [Action] + [Target] + [Context]

Structure: [Action] + [Target] + [Context]

Action Words:

  • Find: Locate, discover, identify
  • Research: Analyze, investigate, study
  • Generate: Create, produce, develop
  • Analyze: Review, examine, evaluate
  • Manage: Organize, track, coordinate

Target Objects:

  • Leads: prospects, contacts, decision makers
  • Companies: businesses, organizations, enterprises
  • Data: information, insights, reports
  • Content: emails, documents, presentations
  • Campaigns: outreach, marketing, follow-up

Context Elements:

  • Industry: SaaS, fintech, e-commerce, manufacturing
  • Location: London, North America, European Union
  • Size: SMB, enterprise, startups, Fortune 500
  • Criteria: verified emails, recent funding, growth signals

Purpose Examples Using the Framework:

  1. Action: Find + Target: qualified leads + Context: in London SaaS companies

    • Result: "Find qualified leads in London SaaS companies"
  2. Action: Research + Target: competitor analysis + Context: for fintech startups

    • Result: "Research competitor analysis for fintech startups"
  3. Action: Generate + Target: follow-up emails + Context: for warm prospects

    • Result: "Generate follow-up emails for warm prospects"

Framework for Writing Effective System Prompts

System Prompt Framework: Role + Context + Instructions + Constraints

Structure: Role + Context + Instructions + Constraints

Role Definition:

Start with "You are a..." and define the agent's professional identity:

  • "You are a B2B lead generation specialist..."
  • "You are a market research analyst..."
  • "You are a sales enablement expert..."

Context Setting:

Provide the business context and environment:

  • "Working with B2B software companies in competitive markets..."
  • "Focusing on early-stage startups seeking growth opportunities..."
  • "Supporting enterprise sales teams with complex sales cycles..."

Instructions:

Specific guidelines for behavior and decision-making:

  • "Always validate contact information before presenting it"
  • "Prioritize companies with recent funding or growth signals"
  • "Use a consultative approach when presenting findings"
  • "Structure responses with clear headings and bullet points"

Constraints:

Limitations and boundaries:

  • "Never make assumptions about contact preferences"
  • "Always verify information through multiple sources"
  • "Respect privacy and compliance requirements"
  • "Focus only on publicly available information"

System Prompt Template:

You are a [ROLE] specializing in [SPECIALIZATION].

Context:
[BUSINESS CONTEXT AND ENVIRONMENT]

Your primary responsibilities:
- [RESPONSIBILITY 1]
- [RESPONSIBILITY 2]
- [RESPONSIBILITY 3]

Communication style:
- [TONE AND APPROACH]
- [FORMATTING PREFERENCES]
- [INTERACTION PATTERNS]

Tool usage guidelines:
- [TOOL 1]: [WHEN AND HOW TO USE]
- [TOOL 2]: [WHEN AND HOW TO USE]
- [TOOL 3]: [WHEN AND HOW TO USE]

Constraints:
- [CONSTRAINT 1]
- [CONSTRAINT 2]
- [CONSTRAINT 3]

How Purpose and System Prompt Work Together

Real-World Example:

Purpose: "Find SaaS companies in London for outbound sales"

System Prompt:

You are a B2B sales research specialist focusing on SaaS companies in the London market.

Context:
You're supporting an outbound sales team targeting mid-market SaaS companies (10-500 employees) in London with proven product-market fit and growth potential.

Your primary responsibilities:
- Identify high-quality SaaS prospects in London
- Verify company information through multiple sources
- Assess company growth signals and funding status
- Compile comprehensive prospect profiles with decision makers

Communication style:
- Use professional, consultative tone
- Present findings in structured formats
- Include confidence levels for data points
- Highlight key opportunities and risks

Tool usage guidelines:
- Business search: Focus on SaaS companies with London offices
- Email validation: Verify all contact information before presenting
- Memory storage: Save important company insights for future reference
- Data export: Structure leads for CRM import with complete profiles

Constraints:
- Only include publicly available information
- Verify company size and location through multiple sources
- Focus on companies with clear SaaS business models
- Respect data privacy and compliance requirements

Result: The Purpose helps users understand this agent finds SaaS leads in London, while the System Prompt ensures the AI behaves like a professional sales researcher with specific guidelines for London SaaS companies.

Best Practices

For Purpose Statements:

  1. Keep it specific: Avoid generic terms like "help with business"
  2. Include key constraints: Location, industry, size, or other criteria
  3. Use action words: Start with verbs like "Find," "Research," "Generate"
  4. Make it scannable: Users should understand it in 2-3 seconds
  5. Align with user goals: Match what users actually want to accomplish

For System Prompts:

  1. Be specific about behavior: Don't assume the AI knows your preferences
  2. Include tool usage guidelines: Specify when and how to use each tool
  3. Set clear constraints: Define what the agent should NOT do
  4. Define success criteria: What makes a good result?
  5. Test and iterate: Refine based on actual agent performance

Common Mistakes to Avoid

Purpose Mistakes:

  • ❌ Too vague: "Help with business development"
  • ❌ Too technical: "Execute lead generation workflows with data validation"
  • ❌ Too long: Multiple sentences or paragraphs
  • ❌ Feature-focused: "Use email tools to find contacts"

System Prompt Mistakes:

  • ❌ Too generic: "Be helpful and professional"
  • ❌ Contradictory instructions: "Be brief but comprehensive"
  • ❌ Missing tool guidance: Not specifying how to use available tools
  • ❌ No constraints: Failing to define boundaries

Next Steps

  1. Practice with templates: Use the frameworks above to create your first agent
  2. Start simple: Begin with clear, specific use cases
  3. Test and refine: Monitor agent performance and adjust prompts
  4. Learn from examples: Study successful agent configurations
  5. Explore advanced features: Multi-channel communication, memory systems, and orchestration tools

Related Articles

Need Help?

If you're having trouble configuring your agents or need personalized guidance, our support team is ready to help you create effective AI agents for your specific use case.

Contact Support | Join Community

We use cookies
Use this dialog to select your cookie preferences. We use cookies to enhance your experience and to collect data for analytics. Read our Cookie Policy.