Getting Started with AI Agents

Learn how to create and configure your first AI agent for lead generation, business research, and workflow automation.

What are AI Agents?

AI agents are specialized artificial intelligence assistants that can perform complex business tasks autonomously. Unlike simple chatbots, DealGate agents can:

• Access powerful tools for lead generation and business research
• Remember information across conversations for continuous learning
• Work across multiple channels (chat, email, webhooks, Slack)
• Execute complex workflows with multiple steps and decisions
• Integrate with external systems through MCP (Model Context Protocol)

Your First Agent: Step-by-Step

Step 1: Access the Agent Dashboard

1. Navigate to Dashboard → AI Agents from your main menu
2. Click "Create New Agent" to start the setup process
3. You'll see a form asking for basic agent information

Step 2: Choose a Clear Agent Name

Pick a name that clearly identifies your agent's purpose:

✅ Good Examples:

• "London SaaS Lead Hunter"
• "Competitor Research Assistant"
• "Email Verification Agent"

❌ Poor Examples:

• "Agent 1"
• "Helper Bot"
• "AI Assistant"

Step 3: Define Your Agent's Purpose

The purpose is what users see when they interact with your agent. Use the [Action] + [Target] + [Context] framework:

Examples:

• Action: Find + Target: qualified leads + Context: in London fintech companies
• Action: Research + Target: competitor analysis + Context: for SaaS pricing strategies
• Action: Validate + Target: email addresses + Context: for existing prospect lists

Your Purpose Should:

• Be specific and actionable
• Include key constraints (location, industry, size)
• Be understandable in 2-3 seconds
• Match what users actually want to accomplish

Step 4: Write Your System Prompt (Optional)

The system prompt controls how your agent behaves. For your first agent, you can start with a simple prompt and refine it later:

Basic Template:

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

Your main goal is to [PRIMARY OBJECTIVE].

Guidelines:
- Use a professional, helpful tone
- Always validate information when possible
- Provide clear, structured responses
- Ask clarifying questions when needed

When presenting results:
- Use bullet points for clarity
- Include confidence levels for data
- Highlight key findings
- Suggest next steps

Example for Lead Generation:

You are a B2B lead generation specialist focusing on SaaS companies.

Your main goal is to find and qualify high-quality leads for outbound sales.

Guidelines:
- Always validate email addresses before presenting them
- Prioritize companies with recent growth signals
- Use a consultative, professional tone
- Focus on companies with 10-500 employees

When presenting results:
- Include company size and funding status
- Highlight key decision makers
- Provide confidence scores for contact information
- Suggest personalized outreach strategies

Step 5: Test Your Agent

1. Start a conversation: Click "Quick Chat" to test your agent
2. Ask simple questions: Start with basic requests to see how it responds
3. Try different scenarios: Test various use cases you plan to use
4. Monitor tool usage: See which tools the agent chooses automatically

Good Test Questions:

• "Find 5 SaaS companies in London with recent funding"
• "Help me validate these email addresses"
• "What can you tell me about [Company Name]?"
• "Create a list of competitors for [Your Company]"

Step 6: Refine Based on Results

Based on your testing, you may want to:

• Adjust the system prompt for better responses
• Clarify the purpose for better user understanding
• Add specific tool preferences if needed
• Include quality criteria for results

Understanding Agent Channels

Each agent can communicate through multiple channels:

Chat Channel (Default)

• Best for: Interactive conversations and testing
• Features: Real-time responses, tool usage visibility
• Use cases: Manual lead research, ad-hoc queries

Email Channel

• Best for: Automated email processing and responses
• Features: Email parsing, automated responses
• Use cases: Lead qualification from inbound emails

Webhook Channel

• Best for: API integrations and automated workflows
• Features: HTTP endpoints, structured data exchange
• Use cases: CRM integration, automated data sync

Slack Channel

• Best for: Team collaboration and notifications
• Features: Slack integration, team notifications
• Use cases: Team alerts, collaborative research

Agent Memory System

Your agent remembers information across conversations:

What Gets Remembered:

• User preferences (industry focus, company size criteria)
• Successful strategies (what worked well in previous searches)
• Important findings (key insights about prospects or markets)
• Configuration details (quality standards, format preferences)

How Memory Works:

1. Automatic storage: Agent saves important information during conversations
2. Semantic search: Finds relevant memories using natural language
3. Importance scoring: Prioritizes more relevant information
4. Continuous learning: Improves over time with more data

Best Practices for New Users

Start Simple

• Begin with one clear use case
• Test with small datasets first
• Gradually increase complexity
• Learn from each interaction

Be Specific

• Provide clear criteria and constraints
• Specify format preferences
• Include quality requirements
• Give examples when possible

Monitor Performance

• Review agent responses regularly
• Provide feedback on results
• Adjust configuration based on performance
• Track tool usage patterns

Iterate and Improve

• Refine system prompts based on results
• Update purpose statements for clarity
• Add new use cases gradually
• Learn from successful configurations

Common First-Time Scenarios

Lead Generation Agent

Purpose: "Find B2B software leads in London with 10-100 employees"

System Prompt Focus:

• Email validation requirements
• Company size verification
• Information quality standards
• Output format preferences

Research Agent

Purpose: "Research competitor pricing and positioning for SaaS companies"

System Prompt Focus:

• Data source preferences
• Analysis depth requirements
• Reporting format
• Update frequency

Validation Agent

Purpose: "Validate and enrich existing prospect lists"

System Prompt Focus:

• Data quality standards
• Validation criteria
• Error handling
• Batch processing preferences

Troubleshooting Common Issues

Agent Doesn't Understand Requests

• Solution: Make requests more specific and include context
• Example: Instead of "find leads," try "find 10 SaaS leads in London with recent funding"

Results Are Too Generic

• Solution: Add specific criteria to your system prompt
• Example: Include company size, industry, location, and quality requirements

Agent Uses Wrong Tools

• Solution: Specify preferred tools in your system prompt
• Example: "Always use email validation tools before presenting contact information"

Inconsistent Response Quality

• Solution: Define clear quality standards and success criteria
• Example: "Include confidence scores for all data points"

Next Steps

Once you're comfortable with your first agent:

1. Explore multi-channel communication for automated workflows
2. Learn about advanced tools for complex research tasks
3. Set up memory management for long-term learning
4. Create specialized agents for different use cases
5. Integrate with external systems using MCP

Related Articles

• Agent Configuration Guide
• Agent Tools & Capabilities Overview
• Multi-Channel Agent Communication
• Agent Memory & Persistence
• Agent Best Practices

Need Help?

Getting started with AI agents can be complex. Our support team is ready to help you create effective agents for your specific use case.

Contact Support | Join Community | Book a Demo