How to Automate Lead Nurturing with Make AI Agents

The recent introduction of the "Make AI Agents (New)" node represents a significant evolution in automated workflows, transitioning away from strictly sequential node-based programming toward autonomous, goal-oriented AI systems. This video masterclass from SupaSaito explores the core functionalities of this powerful new node, demonstrating how it can autonomously select and deploy sub-tools to complete complex objectives based entirely on plain-text instructions. Viewers will discover how to configure an agent's role, define its available data sources, and outline its workflow without building traditional, rigid automation pathways.

To illustrate the practical value of this technology, the video walks through the creation of a fully autonomous sales agent designed for sophisticated lead nurturing. Utilizing Google Sheets as a lightweight CRM database, the AI agent is tasked with reaching out to new newsletter subscribers. The demonstration showcases the agent's ability to seamlessly read lead data, review past communication history, craft personalized introduction emails, and automatically schedule its own future follow-up tasks.

By completing this tutorial, viewers will achieve the capability to build dynamic, context-aware AI systems that drastically reduce manual sales operations. The video highlights advanced scenarios where the agent successfully parses a hypothetical reply from a lead, autonomously searches a "materials" database to find a highly relevant case study, and drafts a tailored response containing the requested information and a consultation booking link. Ultimately, viewers will learn how to deploy AI to handle the initial stages of the lead-to-meeting pipeline, elevating prospect intent before seamlessly handing the relationship over to a human closer.

• Autonomous Sub-Tool Execution: Unlike traditional Make scenarios, the new AI Agent node can independently evaluate a task and select the appropriate sub-tools (e.g., searching records, updating leads, creating tasks) required to complete its objective.
• Plain-Text Workflow Configuration: Complex operational logic is now established through detailed text instructions outlining the agent's persona, available data, and step-by-step workflow, essentially replacing visual node mapping with natural language prompting.
• Comprehensive CRM Automation: The agent is capable of managing end-to-end lead nurturing tasks, including sending initial outreach, logging interaction histories, and proactively scheduling follow-ups for non-responsive leads.
• Dynamic Response Handling: When presented with specific inquiries from leads, the AI agent can dynamically query internal databases to retrieve relevant marketing materials—such as targeted case studies—and seamlessly incorporate them into its replies alongside booking links.
• Transparent "Thinking" and Reporting: The agent generates highly detailed, step-by-step reports of its internal logic and actions, providing operators with full visibility into how it identified the lead, why it chose specific actions, and what materials it selected.

• 2:24 - Opening the AI Agent node to configure the agent's instructions, role, and available data sources.
• 3:10 - Defining the plain-text workflow logic, which replaces traditional visual node mapping with written, step-by-step operational instructions.
• 3:37 - Passing the task name and description as input information for the agent to use before it starts working.
• 3:50 - Adding sub-tools to the agent by hovering over the plus button in the interface.
• 3:59 - Adding the "search lead by email" subnode to filter records.
• 4:13 - Configuring dynamic filter values, allowing the AI agent to decide the value based on the task description rather than manually typing or binding the data.
• 5:04 - Executing a "run once" test to initialize the agent and observe its autonomous tool-calling process.
• 5:32 - Configuring the final node in the main scenario to update the task with a detailed step-by-step execution report generated by the agent.
• 9:35 - Creating a hypothetical customer reply in the database to challenge the agent and test its dynamic response capabilities.
• 10:03 - Executing the scenario again to test the agent's ability to autonomously find relevant materials and handle the newly created lead reply.

Automating Lead Nurturing with the Make AI Agent Node

Overview The Make AI Agents (New) node introduces a paradigm shift in automation, moving from rigid visual workflows to plain-text, goal-oriented instructions. This guide walks you through configuring an autonomous sales agent that manages CRM records, drafts personalized outreach, and dynamically handles customer inquiries.

Step 1: Set Up the Database (0:58 - 1:14)

• How: Create a Google Spreadsheet with distinct sheets for your CRM data. For this workflow, create tabs for Leads, Tasks, History, and Materials.
• Why: Google Sheets currently provides a more reliable and lightweight setup for the Make AI Agent node's dynamic filters compared to alternatives like Notion, which can experience compatibility issues.

Step 2: Configure the AI Agent’s Role and Instructions (2:21 - 2:42)

• How: Open the Make AI Agent node and write a detailed text prompt defining the agent's instructions and persona. For example, instruct it to act as a sales agent aiming to nurture leads and book meetings.
• Why: Establishing a clear persona and setting expectations ensures the agent understands its core objectives and writes communications in the correct tone on your behalf.

Step 3: Define the Plain-Text Workflow (3:10 - 3:35)

• How: In the prompt configuration, write out the step-by-step operational logic the agent must follow, using pure plain text.
• Why: This plain-text instruction essentially replaces the traditional method of stringing together multiple visual nodes in Make, allowing the AI to dynamically process operations one after another.

Step 4: Pass Task Inputs to the Agent (3:37 - 3:44)

• How: Configure the "input" field of the AI agent node to receive the specific task name and task description.
• Why: Passing this data gives the agent the initial context it needs (e.g., the target email address and the desired action) before it begins executing its workflow.

Step 5: Add Sub-Tools (3:50 - 4:10)

• How: Hover over the "+" (plus) button at the bottom of the node interface to add specific sub-tools, such as a "search lead by email" subnode.
• Why: Sub-tools equip the AI agent with the actual functional abilities to read, update, or create records in your database, allowing it to autonomously search leads, log history, and schedule future follow-up tasks.

Step 6: Configure Dynamic Filters (4:12 - 4:29)

• How: Instead of manually typing values or strictly binding data variables into the filter fields, configure the parameters so that the AI agent decides the value based on its input.
• Why: Dynamic filtering empowers the agent to autonomously extract the necessary information (like an email address) from a plain-text task description and apply it to a database search, removing the need for rigid hard-coding.

Step 7: Execute an Initial Test Run (5:03 - 5:28)

• How: Click "Run once" to initiate the scenario.
• Why: Running a test allows you to observe the AI agent "thinking" and verify that it is autonomously calling the correct sequence of sub-tools to complete the task.

Step 8: Update the Task with an Execution Report (5:32 - 5:53)

• How: Configure the final standard node in your main scenario to update your CRM task log with the detailed report output generated by the AI agent.
• Why: The agent generates a highly detailed, step-by-step summary of its internal logic. Logging this ensures full visibility into how the agent identified the lead, why it chose specific actions, and what links it decided to include.

Step 9: Test Dynamic Response Handling (9:22 - 10:22)

• How: Manually create a hypothetical customer reply in your database asking a specific question (e.g., inquiring about Webflow expertise), and run the scenario again.
• Why: This step proves the agent's advanced autonomous capabilities. It forces the agent to read the reply, autonomously search the "Materials" database for a relevant case study, and draft a tailored response containing the requested information alongside a consultation booking link.

Unlike traditional sequential nodes, the Make AI Agent node operates autonomously by using plain-text instructions instead of rigid visual mapping. It intelligently evaluates tasks and dynamically selects the appropriate sub-tools required to achieve its defined objective.

You can build an autonomous sales agent that connects to a CRM database to automatically manage new subscriber outreach. The AI can independently read lead data, draft personalized emails, log interaction history, and proactively schedule future follow-up tasks for non-responsive leads.

Yes, an appropriately configured AI agent can evaluate a lead's email reply and autonomously search internal databases for relevant resources. It then drafts a customized response that seamlessly incorporates targeted materials, such as specific case studies, alongside meeting booking links.

Google Sheets is currently highly recommended for building lightweight CRM prototypes with the Make AI Agent node due to its reliable setup and sub-tool compatibility. While other platforms like Notion can be used, they may currently present technical issues regarding filter configurations.