Article Type: Tutorial
Audience: All Users
Last Updated: January 2026
Overview
This tutorial guides you through creating custom AI marketing agents that transform your team into specialized consultants with immediate access to expert knowledge across digital marketing domains. You'll build agents for Google Ads, SEO, programmatic advertising, and competitive research using curated, domain-specific intelligence.
Learning Objectives
Build AI agents using Agent Builder for rapid deployment
Create comprehensive datastores with expert marketing resources
Configure custom agents with specific instructions and model selection
Deploy and test AI marketing consultants for real-world scenarios
Prerequisites
You'll Need:
Active workspace access with membership in a workspace
Marketing resources and documentation to upload (PDFs, URLs, text files)
Understanding of your team's marketing specialization needs
Basic familiarity with workspace chat functionality
Understand AI Marketing Agent Benefits
Access to Expert Knowledge
Custom AI agents provide your team with a comprehensive library of curated marketing information sourced from authoritative channels. This creates a wide, accurate, and current database that delivers well-informed, precise responses relevant to your specific marketing needs.
Cost-Effective Solution Implementation
Deploying AI marketing agents generates significant ROI through increased efficiency, saving an average of 15-25 hours per week per team member. Marketing outcomes improve with 30-40% faster decision-making. Teams avoid the cost of onboarding multiple channel experts, saving approximately $80K-150K annually per specialized role.
Result: Your marketing team gains enhanced capabilities for information discovery and effectiveness improvement without proportional increases in headcount or training costs.
Build Your Knowledge Foundation
Create a Comprehensive Datastore
Datastores serve as the knowledge base your AI agent references during conversations. The quality and organization of these resources directly impact response accuracy.
Navigate to the left panel and locate the "Datastores" option
Click "Datastores" to access the datastore management interface
Initiate datastore creation
Click the "Create New Datastore" button in the top-right corner of the interface
Assign a descriptive datastore name using the format: [Department] - [Specialization] - [Year]
Example: "Performance Marketing - Google Ads - 2024"
Populate the datastore with resources using one of two methods:File Upload Method:
Drag and drop files directly into the upload area
Supported formats: PDF, DOC, DOCX, TXT, CSV, JSON
Maximum file size: 50MB per file
URL Import Method:
Paste URLs into the URL input field
Separate multiple URLs with commas
Up to 500 URLs per datastore (premium plans support up to 2,000)
Result: Your datastore becomes a comprehensive, searchable knowledge base ready for agent integration.
Tip: Include resources from authoritative sources like Google, Meta, and LinkedIn official documentation; industry publications such as Search Engine Journal, HubSpot, and Moz; and brand guidelines, competitive intelligence reports, campaign performance data, and creative assets.
Configure Your Custom Marketing Agent
Access Agent Creation Interface
Navigate to the left panel and click the "Agents" option
This opens the agents management interface
Organize your agent in a category
Select an existing category from the dropdown, or
Create a new category by clicking "Create Category" and entering a name
Use the "Marketing Consultation" folder if available
Initiate agent creation
Click the "Create Agent" button (top-right corner)
Define Agent Identity and Core Configuration
Configure basic agent information
Enter Title: Assign a clear, descriptive agent name that identifies its specialization
Write Description: Create a summary visible during chat interactions that explains the agent's purpose
Specify agent purpose and function with detailed guidanceDefine Primary Goal
State the core mission clearly (e.g., "Generate qualified leads for enterprise software," "Create engaging social content that drives website traffic")
Use action-oriented language that defines measurable outcomes
Identify Target Audience
Describe your ideal client with specific details:
Demographics: industry, company size, role, experience level
Pain points and challenges they face
Decision-making authority and budget parameters
Communication preferences and tone expectations
Example: "Mid-level marketing managers at B2B SaaS companies (50-500 employees) struggling with lead quality, seeking data-driven solutions"
Specify Content Format Requirements
Define deliverable types and constraints:
Format: Social media posts, blog articles, email campaigns, sales scripts, video scripts
Length: Character limits, word counts, duration
Style: Professional, conversational, technical, playful
Brand Guidelines: Voice, tone, terminology, compliance requirements
Establish Specific Behaviors & Requirements
Define response structure (bullet points, paragraphs, templates)
List required elements to always include (CTAs, disclaimers, data points)
Identify elements to avoid (jargon, competitor mentions, sensitive topics)
Set quality standards and approval workflows
Example: "Always include 3-5 bullet-point USPs, end with a qualifying question, cite data sources, and maintain brand voice"
Result: The Agent Builder will generate a complete instruction set based on your specifications. Review the AI persona section, primary goal description, required inputs, critical policies, and identify which tools (web search, keyword research, etc.) will be needed.
Select AI Model and Datastore Connection
Choose the appropriate AI model based on your requirements
Model Selection Criteria | GPT-5.2 | GPT-4o-Mini |
Dataset Size | Best for extensive datasets (10,000+ resources) | Optimized for smaller datasets (under 10,000 resources) |
Response Depth | Comprehensive, detailed analysis | Quick, concise responses |
Processing Speed | 3-5 seconds average response | 1-2 seconds average response |
Use Case | Complex marketing strategies, multi-step analysis | Routine queries, quick consultations |
Recommendation | Select for strategic planning and detailed client consultations | Select for real-time call support and rapid Q&A |
2. Connect your datastore to the agent
In the "Connect Datastores" section, select the datastore you created in Section 2 from the dropdown list
Click "Continue" to enable the agent's access to your curated knowledge base
3. Review configuration summary
Verify agent title, description, and instructions are accurate
Confirm model selection aligns with your speed vs. depth requirements
Ensure datastore connection is active and accessible
4. Publish your agent
Click the "Publish agent" button to activate your agent
Wait for the confirmation message indicating successful deployment
Result: Your AI marketing agent is now fully functional with access to your knowledge base and ready for real-world deployment.
Test and Deploy Your Agent
Verify Agent Functionality
Access the chat interface
Navigate to the "Chat" section in the left panel
Select your newly created agent from the agent dropdown menu
Conduct comprehensive testing
Ask 5-10 varied marketing questions covering your specialization
Test questions should include:
Simple factual queries (e.g., "What are the latest Google Ads character limits?")
Complex strategic questions (e.g., "Develop a Q1 campaign strategy for SaaS lead generation")
Edge cases outside your datastore content
Evaluate response quality
Verify accuracy against your source materials
Assess relevance to your specified target audience
Confirm adherence to content format requirements
Check for inclusion of required elements (CTAs, data citations, etc.)
Refine agent instructions based on test results
Return to the Agents section and select your agent
Click "Edit" to modify instructions, datastore connections, or model selection
Re-publish the agent after making adjustments
Result: A validated AI marketing consultant ready for team deployment with confidence in its accuracy and usefulness.
Deploy for Real-World Scenarios
Implementation Use Cases:
Digital marketing trend research and analysis
Client call preparation and real-time support during meetings
Campaign strategy development and refinement
Consumer behavior analysis and insights
Competitive intelligence gathering
Share the agent with your team
In the agent settings, locate the "Sharing" or "Permissions" section
Add team members by email or workspace groups
Set appropriate access levels (view, edit, admin)
Integrate into workflows
Use the agent during client interactions for real-time consultation
Incorporate agent queries into campaign planning meetings
Establish standard operating procedures for when and how to engage the agent
Result: Enhanced marketing capabilities with immediate access to expert-level consultation across your entire team.
Best Practices
Curate high-quality resources: Include only verified, authoritative sources in your datastore to ensure accurate responses. Regularly audit and remove outdated or low-quality content.
Write specific, detailed instructions: Provide comprehensive behavioral guidelines to ensure consistent, useful agent responses. Include examples of good vs. bad outputs.
Test continuously with new scenarios: Regularly evaluate your agent with real-world queries and refine instructions based on performance gaps.
Keep datastores current: Schedule quarterly reviews to update resources and maintain relevance. Remove deprecated platform information and add new best practices.
Share strategically with access controls: Deploy agents across your team while maintaining appropriate access controls for sensitive competitive intelligence and client data.
Document your agent's capabilities: Create a one-page reference guide for each agent that outlines its specialization, best use cases, and example prompts to help team members maximize its value.
Monitor usage and performance: Track which agents are used most frequently and gather feedback from team members to identify improvement opportunities and additional specialization needs.
Common Questions
Q: How many resources can I include in a datastore?
A: Datastores can typically handle extensive resource collections, but performance may vary based on your plan level and total content volume. Most users maintain optimal performance with 500-1,000 high-quality resources.
Q: Can I modify agent instructions after creation?
A: Yes, you can edit agent instructions, descriptions, and connected datastores at any time. Changes take effect immediately upon re-publishing the agent.
Q: Which AI model should I choose for marketing tasks?
A: Select GPT-5.2 for comprehensive analysis and complex marketing strategies requiring detailed responses. Choose GPT-4o-Mini for quick responses and routine queries where speed is prioritized over depth.
Q: Can multiple team members use the same agent simultaneously?
A: Yes, agents can be shared across your workspace, allowing multiple team members to access the same marketing expertise concurrently without performance degradation.
Q: How do I ensure my agent provides accurate information?
A: Curate high-quality source materials, test thoroughly with various scenarios (including edge cases), regularly update your datastore with current information, and implement a feedback loop where team members report inaccuracies.
Q: Can I integrate external marketing tools with my agent?
A: Advanced users can create custom plugins to integrate databases, CRM systems, and API endpoints for enhanced functionality. Contact your workspace administrator or consult the plugins documentation for implementation guidance.
Q: What happens if my agent cannot find relevant information?
A: The agent will transparently indicate when it cannot find specific information and may suggest alternative approaches, request clarification, or recommend consulting a human expert for highly specialized scenarios.
Q: How do I organize multiple marketing agents for different specializations?
A: Create separate agents for distinct marketing domains (SEO, PPC, Social Media) with specialized datastores and instructions for each area. Use clear naming conventions and categorize agents logically for easy team navigation.
Q: Can I restrict which datastores an agent can access?
A: Yes, you control datastore connections at the agent level. An agent only accesses explicitly connected datastores, allowing you to segment sensitive or specialized information across different agents.
Q: How often should I update my marketing agent's datastore?
A: Update quarterly at minimum, or immediately when major platform changes occur (Google algorithm updates, new ad formats, policy changes). Set calendar reminders to review and refresh content regularly.