Learning Timeline
Key Insights
Context Size Impact
Setting the Context Size to 'High' is critical for research agents. It allows the AI to retain more information found during web searches before summarizing it, resulting in more comprehensive lead details.
Prompt Specificity for Data Structure
When writing the system prompt, explicitly listing the Google Sheet column names (e.g., 'Business Type', 'Address') ensures the AI searches for and extracts exactly the data points needed for the final database, rather than generic summaries.
Prompts
Lead Finder System Instructions
Target:
OpenAI Agent Builder (System Prompt Field)
You are a helpful lead generation agent. Your goal is to find leads based on the user's search criteria.
You must find the following specific information for every lead to match our database columns:
1. Name
2. Business Type
3. Address
4. Phone Number
5. Email
6. Description
7. What makes this business different
Once you have gathered this information via Web Search, format it as text and pass this information to the next agent (Data Entry Agent).
Step by Step
Creating the Lead Database Template
- Open a new blank Google Sheet.
- In the first row (Header Row), create the following columns:
- Column A: Name
- Column B: Business Type
- Column C: Address
- Column D: Phone Number
- Column E: Email
- Column F: Description
- Column G: Differentiation (What makes the business different)
- Column H: Email Sent Status
- Name the Google Sheet 'Lead Database' (or similar) for easy reference.
Initializing the Lead Finder Agent
- Open your Agent Builder interface and click 'Create' to start a new project.
- Locate the default nodes on the canvas. If unnecessary nodes exist, select them and press Delete/Backspace to clear the workspace.
- Ensure a 'Start' node is present on the canvas.
- Drag and drop an 'Agent' node from the sidebar onto the canvas.
- Click and drag from the output handle of the 'Start' node to the input handle of the new 'Agent' node to connect them.
- Click on the new 'Agent' node to open its configuration settings.
- In the 'Name' field, type 'Lead Finder Agent'.
Configuring Agent Instructions and Tools
- In the Agent configuration panel, locate the 'System Instructions' or 'Prompt' field.
- Paste the system prompt defined in the 'System Instructions' card below.
- Toggle the 'Chat History' switch to ON (this enables conversational memory).
- Select the desired AI Model from the dropdown (e.g., GPT-4o or GPT-5).
- Navigate to the 'Tools' or 'MCP Server' section.
- Select and enable 'Web Search' to allow the agent to browse the live internet.
- Locate the 'Context Size' setting and change it to 'High' to allow processing of large amounts of search data.
- Set the 'Output Format' to 'Text'.
- Click 'Save' or 'Play' to finalize the agent setup.
