How to Build AI Agents in Make.com (Complete 2025 Guide)
Struggling to keep up with repetitive tasks that require decision-making? Discover how intelligent AI agents in Make.com can manage your workflows, update databases, and automate complex processes - all through natural language commands. Transform from overwhelmed business owner to automation leader in one tutorial.
AI Agents vs Regular Automation: Key Differences
Most business owners start with basic automations - if X happens, then do Y. These work well for simple, repetitive tasks but fall short when decisions or adaptations are needed. That's where AI agents change the game.
While regular automations follow fixed Boolean logic with predetermined steps, AI agents use fuzzy logic to dynamically decide which tools to use and how many times to use them. An AI model like ChatGPT steers the entire process, allowing your automation to adapt based on context and input.
Key insight: Use regular automations for predictable, linear processes (like saving form submissions to a CRM). Use AI agents when you need intelligent decision-making, context awareness, or the ability to handle variable workflows.
Real-World Example: Task Manager Agent
The tutorial demonstrates a powerful task manager agent that interacts with Notion via Telegram. Every morning at 8:30 AM, it sends a list of open tasks. Throughout the day, you can message commands like:
- "Mark the script writing task as completed"
- "Move all due dates for next week to the following week"
- "Create a new task for tomorrow to design the video thumbnail"
The agent handles these natural language requests, intelligently determining which tools to use (get tasks, update tasks, create tasks) and executing the changes in Notion automatically. This solves the common problem of forgetting to update task statuses or letting due dates slip.
Make.com Interface Tour
Make.com organizes AI agents separately from regular automations (called scenarios). The left menu has two key sections:
- Scenarios: Where you build traditional automations with fixed steps
- AI Agents: Your control center for intelligent agents that can make decisions
When creating a new agent, you first select which AI model to use (OpenAI, Gemini, or Anthropic) and connect it via API key. Make provides direct links to each provider's API key generation page for easy setup.
Setting Up Your First AI Agent
Creating an agent begins with basic configuration:
- Name your agent (e.g., "Task Manager Agent")
- Select your AI model (GPT-4 recommended for complex agents)
- Write an initial system prompt (we'll refine this later)
The agent configuration page is where you'll train and build your AI agent by adding tools. You can immediately test basic functionality in the built-in chat interface before connecting it to external apps.
Pro tip: Start with a simple system prompt like "You are a helpful assistant" for initial testing, then progressively add complexity as you build out the agent's capabilities.
Connecting Notion to Your Agent
To give your agent access to Notion, you'll need to:
- Create an internal integration in Notion (under Settings → Connections)
- Generate an integration token
- Add the connection in Make.com's Notion module settings
- Grant the integration access to specific databases
A clever trick shown in the tutorial: Create a temporary formula field in Notion to display your user ID, which you'll need for filtering tasks assigned specifically to you. This ensures your agent only interacts with your tasks, not your team's.
Building Agent Tools: Get, Create & Update Tasks
The task manager agent uses three core tools:
- Get Open Tasks: Searches your Notion task database for items with status="open" and assigned to you
- Create New Task: Adds items to your Notion database with dynamic fields (name, due date) and fixed fields (assigned to you)
- Update Task: Modifies existing tasks (status, due date, etc.) using a special scenario workaround
The tutorial reveals an important limitation: Currently, you can't directly let the agent dynamically decide which database item to update in a module tool. The workaround is to create a separate scenario that your agent can run, passing the item ID as a parameter.
Crafting the Perfect System Prompt
As you add tools, your system prompt needs to evolve. The tutorial provides a detailed prompt that:
- Defines the agent's role (personal task management assistant)
- Lists available tools and their purposes
- Provides specific instructions for using each tool
- Specifies response formatting (e.g., hyperlinked task names)
You can enhance this with dynamic variables like the current date and time, helping your agent understand temporal references ("move to next week"). The full system prompt is available in the tutorial resources.
Telegram Integration for Mobile Access
To interact with your agent via Telegram:
- Create a Telegram bot using @BotFather
- Copy the API token into Make.com's Telegram module
- Set up a scenario that watches for Telegram messages
- Add filters to ensure only your messages trigger the agent
- Connect the Telegram input to your AI agent
- Configure the agent to reply back to Telegram
This creates a seamless mobile interface where you can manage tasks through natural language commands, whether you're at your desk or on the go.
Setting Up Scheduled Agent Tasks
The tutorial shows how to create a second scenario that runs your agent on a schedule (weekdays at 8:30 AM) to:
- Automatically message you your open tasks each morning
- Use the same Telegram chat thread for consistency
- Maintain all the agent's capabilities through one interface
This demonstrates the power of Make.com's AI agents - once built, you can trigger them through multiple channels (scheduled, message-based, or from other automations) while maintaining a single, intelligent backend.
Watch the Full Tutorial
See the complete 39-minute walkthrough of building this AI agent system, including timestamped sections on connecting Notion (12:45), setting up the Telegram integration (28:10), and crafting the perfect system prompt (22:30).
Key Takeaways
AI agents in Make.com represent the next evolution of business automation, moving beyond rigid if-then logic to intelligent, adaptable systems that understand context and make decisions. The task manager agent demonstrated in this tutorial solves a universal pain point - keeping task systems updated - while showcasing techniques applicable to countless other business processes.
In summary: 1) Use AI agents when you need decision-making in your workflows, 2) Connect them to your core business tools like Notion, 3) Interact through natural language interfaces like Telegram, and 4) Schedule them for proactive assistance. This combination delivers what every busy business owner needs - an intelligent assistant that handles the mundane so you can focus on growth.
Frequently Asked Questions
Common questions about this topic
Regular automations follow fixed Boolean logic with predetermined steps, while AI agents use fuzzy logic to dynamically decide which tools to use and how many times to use them. AI agents adapt based on input and have an AI model like ChatGPT steering the entire process.
For example, a regular automation might always save form submissions to your CRM in the same way. An AI agent could examine each submission, research the company, and decide whether to create a new record or update an existing one based on what it finds.
- Regular automation: Fixed path, no AI, predictable outcomes
- AI workflow: Fixed path with AI in some steps
- AI agent: Dynamic path with AI steering the entire process
AI agents can manage task lists in Notion, research leads automatically, classify incoming emails, update CRM records intelligently, and handle complex multi-step workflows that require decision-making. The tutorial demonstrates a task manager agent that updates Notion tasks via Telegram.
Other powerful use cases include customer support triage (classifying and routing tickets), intelligent document processing (extracting and acting on key information), and dynamic content generation (creating personalized marketing materials based on customer data).
- Lead research agents that build prospect dossiers
- Email processing agents that categorize and prioritize
- Data analysis agents that spot trends and alert you
No coding is required. Make.com provides a visual interface where you connect modules representing different tools and actions. The platform handles all the underlying complexity, allowing business users to create sophisticated agents through configuration rather than programming.
That said, some technical comfort helps when working with API keys, understanding data structures, and troubleshooting connections. The tutorial walks through each step visually, showing exactly where to paste tokens and how to configure modules without writing any code.
- Visual interface with drag-and-drop modules
- Pre-built connectors for popular apps
- Configuration-based rather than code-based
Make.com supports OpenAI models (like GPT-4), Google's Gemini, and Anthropic's Claude. You connect these via API keys in the agent settings. The platform provides direct links to each provider's API key generation page for easy setup.
The choice depends on your needs - GPT-4 offers strong reasoning for complex agents, Claude excels at following instructions precisely, and Gemini provides good value. You can test different models to see which works best for your specific use case.
- OpenAI (GPT-3.5, GPT-4)
- Google Gemini
- Anthropic Claude
You create an internal integration in Notion to generate an API token, then connect it in Make.com's Notion module settings. The tutorial shows how to set up specific permissions so your agent can only access designated databases while maintaining security.
A key step is granting the integration access to each database you want to use. The tutorial includes a clever trick using a temporary formula field to find your Notion user ID, which you'll need for filtering tasks assigned specifically to you.
- Create Notion internal integration
- Generate and copy API token
- Connect in Make.com with proper database permissions
Yes, one powerful feature is that agents can execute complete scenarios as tools. This lets you package multi-step automations that your agent can trigger intelligently based on conversational input, creating sophisticated chained workflows.
The tutorial demonstrates this with the task update functionality - since module tools couldn't dynamically select database items, a scenario was created that the agent runs when needed, passing the item ID as a parameter.
- Package complex workflows as scenarios
- Agents can run them with dynamic inputs
- Combine the power of both approaches
The tutorial demonstrates setting up Telegram integration, allowing you to message your agent naturally from mobile or desktop. You can also schedule your agent to run automatically (like a morning task reminder) or trigger it from other apps via webhooks.
Telegram works particularly well because it's available across devices, supports rich formatting, and maintains conversation history. The tutorial shows how to set up filters so only your messages trigger the agent, preventing accidental activations.
- Telegram for natural language interaction
- Scheduled runs for proactive assistance
- Webhooks for triggering from other systems
GrowwStacks specializes in building custom AI agents and automations tailored to your business workflows. Our team can design, implement, and maintain Make.com agents that handle your specific operational challenges, from CRM updates to complex decision workflows.
We offer free consultations to discuss your automation goals and create a solution that saves you hours each week. Whether you need a simple task manager like the one in this tutorial or a sophisticated agent handling multiple business functions, we can build it for you.
- Custom AI agents built for your exact needs
- Integration with your existing tools and workflows
- Free 30-minute consultation to discuss your requirements
Ready to Transform Your Business with AI Agents?
Don't let manual processes and forgotten tasks hold your business back. The GrowwStacks team can build your custom Make.com AI agent in as little as 2 weeks, delivering an intelligent assistant that handles the work you dread.