We have all been there. Your to-do list is a mess, tasks never stop coming, and your backlog is FULL.
It is a Friday night. You are working on the most BORING task you could be working on. And you ask yourself,
“Is this a waste of time?”
In this article, I will show you how you can build your own AI agents. So you can equip ChatGPT with the right tools to automate your tasks.
Table of Contents
The Biggest Misconception
Okay, confession time:
You have probably played with ChatGPT, so you already know how powerful it is.
Here is the kicker—ChatGPT will not release a new version that will suddenly automate all your tasks! You can’t wait for the next update to roll on and do the job for you. You have to take the leash onto your hands.
You need to start building your own AI Agent.
When people hear AI agents, they think of robots or complicated tech jargon.
Relax!
It is just a targeted way of running the big models. You can give it tools, explain the task in small steps and it will go and automate a SINGLE task for you.
YOU NEED AN ARMY OF AGENTS
You need not only one, not two, but an army of these small agents. So, task by task they can automate your business.
“Why do you have to waste time writing the same emails, the same messages, and repetitive tasks?”
An AI agent could have done it for you!
In the next section, I will show you exactly how you can create your AI agent in a few steps—no skills required and no coding degree needed.
How to Create an AI Agent
Let’s get started and build an AI agent together. Don’t worry, we’re not building a ROCKET. As a starter, we will keep it small and create an agent that generates fictional customer complaints.
Step 1: Set up n8n
Head over to n8n. Create a new workflow and add the “Click Trigger”.
Step 2: Add the AI Agent
Click the “plus” icon and drag and drop the “AI Agent” node.
What is this node?
This block is one of the CORE blocks you have to know. And in Step 4 we will go into more detail about.
There are many types of agents n8n supports but for this example, we’re using a “Tool Agent” in the “conversation” mode.
You can think of it as a specialized ChatGpt with the bare minimum tools and customizability.
Step 3: Set Up Memory
For all agents, you need to specify a place to store the chat memory. In most cases, you will use the “Window Buffer Memory”. It enables the agent to store the chat memory in the browser window.
Drag and drop it. Make the “Session ID” “define below” and give it a “key”.
The key can be anything you want it doesn’t matter.
You are all set and there is only 2 more steps to go. Easy peasy!
Step 4: Customize It
This is where you turn this agent into your taskmaster.
Your AI agent can:
- Take instructions (aka prompts).
- Store memory.
- Access tools.
- Access other workflows.
- Access a Vector database.
- Access different methods.
- Output with a specific format.
And can do many more things. But for now, we are going to keep it simple.
We’ll give it the following prompt:
“You’re an expert at writing fictional customer complaints. Create short emails where a customer complains about bad service from a real estate agency. Keep it simple. Max 134 words.”
Step 5: Run It!
Hit that shiny “Test Step” button and watch the magic happen. Here’s what our agent came out without:
Subject: Disappointing Service from Your Agency
Dear [Agency Name],
I cannot express enough my frustration due to the lack of professional service from your agency. Despite being promised seamless assistance in our hunt for a home, we have been met with unresponsiveness and delays. I find it unacceptable to be kept waiting for responses and updates about serious inquiries. As a customer, I expected better from your reputed agency. Kindly rectify this situation.
Sincerely,
[Customer’s Name]
Outro
Boom! In less than a minute, you’ve created an agent that creates complaint mails automatically.
What am I suppose to do with an imaginary complaint mail?
Nothing!!!
This was just a small introduction to show you what you can do with these agents. AI agents are here. They’re small, specialized, and super effective. You can start small—automate one task, then build on that.
And that is exactly what we are going to be doing. In our next article, we’ll level up this agent by connecting it to Gmail so rather than creating imaginary complaint mails it will respond to them.
If you need any help, feel free to redeem your AI strategy call here. And we will take a look at your current setup for you. Until the next episode take care.