Unit Introduction
Welcome to the second unit of the AI, AI agents and automation!
You have just learned about the three types of automation in Make, now it’s time to focus on agentic automation.
You will learn: what you can do with agentic automation
when you should use it
examples of how it can be applied Let’s start!
Use agentic automation
Agentic automation relies on AI agents being able to reach a goal on their own, without step-by-step instructions or human intervention. Agentic automation is best for complex or unpredictable tasks where you can’t
plan every step ahead of time. These situations need smart decisions, flexibility, and the ability to adjust based on what’s happening, not just following fixed rules.
What can you do with
agentic automation?
1 Handle complex rule-based tasks
2 Adapt to different user requests
3 Work with unstructured data
4 Make real-time decisions
5 Use human-like reasoning
Let’s have a look at each of them. 1: Handle complex rule-based tasks
Handle complex rule-based tasks
Some processes involve lots of moving parts: multiple conditions, exceptions,
and specific pieces of information, like numbers, dates, or categories, that change depending on the data. You can build a traditional automation scenario for this, using quite a lot of filters and routes to take into account all the conditions.
These tasks require handling multiple changing factors at once and making decisions based on different pieces of information to direct to the desired outcome.
Your company runs a customer support system that receives tickets with multiple details: the customer’s subscription level, issue type, time of day, support rep availability, and history of past interactions.
You need to assign each ticket to the appropriate support rep based on these changing details. To do this, you need to use filters and routes to take every condition into account.
These processes follow Boolean logic, so traditional automation can handle them, but they often require complex scenarios. AI agents, however, can handle this complexity without needing you to define every single condition in advance. You tell the AI agent the goal and it looks at all the details involved and figures out what needs to be done step by step.
You don’t need to create routes and filters for every condition because the AI agent understands the data and adapts its decisions automatically based on the information it receives.
You want the AI agent to assign each support ticket to the most suitable support rep.
The AI agent analyzes all details, subscription level, issue type, time of day, and agent availability. It evaluates which support rep best matches the ticket’s needs and assigns the ticket accordingly, without requiring you to plan every condition in advance.
2:Adapttodiffe rentusersrequests
Adapt to different users requests
Users can request different things from the same starting information. For example, they might ask for a list of contacts filtered by location, or they might want that same data exported in a specific file format. Traditional automation can handle these varying requests, but it might require creating a separate scenario for each type.
These are tasks where the system needs to handle a variety of user queries that may ask for different types of information or actions related to the same topic. Your company stores all customer contacts in HubSpot. The sales team often requests information from this database.
For example, one salesperson might ask for “potential customers in Madrid” to prepare a call list, while another asks to “export all European leads in Excel” for marketing. Each request needs a different response and format, so you create a separate scenario for each.
This can be handled with traditional automation by creating separate scenarios for each request type. However, as requests grow, the number of scenarios multiplies and becomes difficult to maintain. AI agents, however, can deal with a wide range of requests from the same starting data and adapt their responses automatically, without needing separate scenarios for each variation.
You tell the AI agent the goals to achieve and give it the tools to handle different types of requests. The AI agent receives the requests, interprets their intent, and responds accordingly, without needing separate pre-built scenarios for each case.
You want the AI agent to handle customer data requests.
When one salesperson asks for “potential customers in Madrid” and another requests “export all European leads in Excel,” the AI agent automatically filters, formats, and delivers the right data for each request in the same scenario.
Work with unstructured data Unstructured data comes in many forms like PDFs, images, emails, or text
documents without a fixed structure. Traditional automation requires structured inputs and can only process these kinds of data by using complex, rigid extraction rules designed for specific formats. AI automation can handle unstructured data. However, you still need complex rules or filters because the extracted information can vary widely in type and format.
These tasks involve interpreting varied and messy data formats, extracting useful information, and taking appropriate actions based on that data, even when it is incomplete or irregular.
Your company handles thousands of contracts daily, stored as PDFs from clients and partners. When a contract is uploaded, you need to extract all the information, check if it’s complete, and either approve it or request the missing data.
This could be an AI automation case. You use AI modules to extract information from each contract PDF. Then, you need to set up routes and filters to handle the extracted data and cover every possible condition. This approach works, but can get complex as you need to plan all the different conditions. AI agents, however, can interpret the data more flexibly and make decisions based on the data extracted without requiring you to build detailed rules for every case.
You tell the AI agent the goal.
The AI agent reads and understands the document’s content and decides what actions to take without relying on fixed rules for every possible data case.
You want the AI agent to review the contracts, ensure all required data is included, and request any missing information. When a user uploads a contract, the AI agent reads the document and understands its content. It checks that all the needed information is present, and takes the appropriate action, either approving the contract or requesting additional details, without needing predefined rules for every condition.
4:Make real-timedecisions
Make real-time decisions
Many processes require quick decisions based on changing information. Traditional automation follows fixed rules, so it struggles with the unexpected. Even AI automations rely on preset rules and can’t fully react in real time. When things don’t go as planned, people often need to step in and manually perform tasks, because rules can’t cover everything.
These tasks need dynamic decision making, where actions depend on real-time information rather than preset instructions.
Your company uses a single supplier to fulfill orders, and this process is automated. However, if that supplier runs out of stock, the system can’t automatically switch to another supplier or update the order. In this case, manual intervention is needed to find alternatives, adjust the order, and inform the customer.
This is tricky to handle with both traditional automation and AI automation. To do so, you would need to account for every possible exception and plan for every situation that could arise. This quickly becomes complex and often isn’t worth the effort. Manual intervention can solve the issue more efficiently in many cases. With AI agents, you can rely on them to assess the situation in real time and decide the best action without needing exhaustive pre-planning. You tell the AI agent the goal.
The AI agent assesses the current situation, makes decisions in real- time, and adapts its actions to meet the goal without needing you to specify every step. It can choose the best option based on available information, handle unexpected changes smoothly, and adjust its plan as new data arrives, allowing continuous, flexible problem solving without rigid rules.
You want the AI agent to manage orders. If one supplier is out of stock, the AI agent finds another supplier, updates the order accordingly, and notifies the customer. All automatically, without fixed steps for each case.
5 : Usehuman-like reasoning
Use human-like reasoning
Many tasks require understanding broad, complex goals. Traditional automation can’t handle them well because it needs fixed rules and step-by-step
instructions. AI automation is better at reading inputs but still relies on preset workflows and scenarios. Both struggle with open-ended requests, so they’re not ideal for tasks that need flexible understanding and real-time decisions.
These tasks need an ability to break down high-level goals into smaller actionable steps, make plans, and handle missing information flexibly, much like a human would. Your company is preparing for a product launch next week and needs to coordinate many tasks. To be able to automate it, you need to break down the launch into detailed, specific steps, such as creating timelines, drafting social media posts, checking resource availability, and building workflows for each task. This requires lots of planning and manual work on your part.
Traditional automation needs a detailed breakdown to function, as it relies on fixed steps and rules. Even AI automation needs significant setup to cover every part of the process. AI agents, however, can understand the overall goal and figure out the necessary steps on their own.
You tell the AI agent the overall goal.
The AI agent uses human-like reasoning to break the goal into subtasks, plan their execution, and perform the required actions without needing every step spelled out. You want the AI agent to prepare for a product launch next week. The AI agent creates a timeline, drafts social posts, checks if the product page exists, and requests missing info, all autonomously and adapting to changing circumstances.
When should you
use agentic automation?
Use agentic automation when the process involves unpredictable data, exceptions, or changing conditions, and you want the system to adapt and decide what to do without predefining every step.
When designing your workflow, you might wonder which type of automation best fits your needs. Here’s a quick overview of the key characteristics of each. Traditional automation
You define every step manually
It follows fixed rules
AI automation
You define every step manually
You can add smart capabilities (e.g., document parsing, classification) to fixed workflows
It relies on pre-built routes and structure Agentic automation
You define the goal, not the steps
The AI agent figures out the path dynamically, adapting to different conditions
Here are some checklists that can help you figure out which kind of automation your project needs. Traditional automation
The tasks are very repetitive and well-defined with clear, fixed steps.
All possible user requests or scenarios are easy to list upfront.
Data inputs follow strict, structured formats.
The process requires simple, rule-based decisions without much variability.
The number of variations or exceptions is limited.
You’re fine with manually updating the automation whenever new cases appear.
AI automation
The task involves some variation in input data but remains mostly predictable.
AI can help understand the data, but you still need to set up clear steps for each case.
Inputs are semi-structured or include some unstructured content, like text or scanned documents. You can define rules or filters to handle most expected outcomes.
Manual intervention is acceptable for exceptions.
Agentic automation
The task involves diverse and unstructured inputs or open-ended user requests.
The system needs to interpret high-level goals and decide dynamically what to do.
Flexibility is key, especially in case of unexpected situations.
You want the system to reason, plan, and adapt like a human, not follow rigid workflows.
You aim to reduce manual rule creation and ongoing maintenance for changing cases.
You need to make decisions in real time and deal with multi-step problems. If you still have doubts whether to use AI automation or agentic automation, follow this rule:
AI automation = You plan the path, AI helps with tasks.
Agentic automation= You give the goal, the AI agent plans the path.
Examples
Here are some examples of AI agents that you can build with Make.
AI recruiting agent
CHECK IT OUT
Sales outreach AI agent
CHECK IT OUT In the next course, you’ll learn how to build a simple AI agent yourself.
Continue to the wrap up of this unit
Wrap up
Agentic automation uses AI agents that can think, adapt, and make their own decisions to reach a goal. Instead of following fixed rules, the AI agent decides what steps to take based on the situation. It’s great for complex tasks where things change often or can’t be fully predicted. This kind of automation works well with changing conditions, figuring things out, learning from new information, and adjusting as needed.
In Make, you can use AI agents in your scenarios for several purposes. They can handle complex rules, make real-time decisions, adapt to different user requests, and work with unstructured data like PDFs or images. They can also reason like a human to figure out the best way to reach a goal, without needing to define every step in advance.
Your process requires AI automation when you plan a clear sequence of steps and you want to enhance specific tasks with AI capabilities. In contrast, agentic automation is better suited when you define the overall goal and let the AI agent independently determine the best path to achieve it, adapting dynamically as needed.
You’ve completed the second unit!
Amazing job!
You now know when to use AI agents in your automations. Let’s keep this momentum going, and move on. In the next course
you will learn how to build and use your AI agents in Make.
Mark this task complete to continue to the next unit.