How We Automate Our Own Operations With n8n + AI Agents

Completely automating all my daily tasks is one of my biggest career goals. The idea of having an “army” of “independent” AI agents that work together to achieve my goals is simply too enticing.

But so far, every attempt at automating marketing and growth efforts has shown me that some things are better left unautomated.

Unless I could achieve a level of quality that I’m really satisfied with.

I’m still far from automating all my work (there’s always going to be more work 😄), but it’s finally beginning to sound plausible.

Today, I’ll walk you through some of my thoughts on automations and show you how we help clients (and ourselves) automate boring repetitive tasks — and make sure they’re all executed with a great level of quality and precision.

Why even bother automating?

It’s no secret that automation tools have been available for quite some time. From Zapier to Make.com and now n8n, they’ve been evolving and quietly solving our biggest pain point — the lack of time.

It’s the lack of time to do meaningful things that move the needle. Just ask yourself — how much of this week’s work was something that really matters? How much of that time was spent on entering data? Or copying data from one place to another? Or generating yet another document and sending it through?

This gruntwork rarely carries any real value. And it’s a real thief, if I may say. It’s been stealing a few hours/week from you — for years.

You see what I’m aiming at with this?

Nevertheless, we are plagued with this “it’s just 5-minutes” syndrome.

“Why spend hours automating something when it only takes 5 minutes to do it?”

Well, if you do that 5-minute task 5 times a week — it’s 25 minutes now!

And what if you’re not alone? What if 10 of your colleagues have to do the same? That’s 250 minutes now! Just over 4 hours a week - if you and your 10 colleagues have a 5-minute task to do every day.

Can we do something about that?

How to spot tasks that are ripe for automation?

The answer is pretty simple. If it’s a task that doesn’t require a lot of brainpower, there’s a chance it can, and it should, be automated. If you have a recurring task, you should always think about automating it.

If it's possible, of course. (It almost always is!)

When approaching automation, there are a few things to think about:

• What do I want to automate and why?
• How much time is this going to save me?
• Which tools do I need?

Of course, the first question being what and why — this is the core one. You’ll want to go over your daily and weekly recurring activities and think: Can I automate that in any way?

For example, there may be a report you need to create and send to stakeholders each week. And that single report may be costing you 15-25 minutes every week. Not anymore!

Finally, let’s see the tools. There is a ton of them out in the wild, but I’ll put only 3 in the spotlight — I think they really deserve it:

1. Zapier
2. Make.com
3. n8n

We’ve all heard of Zapier (and probably used it too). It’s an automation powerhouse designed with non-dev users in mind. Awesome if you want to connect an app or two and design a simple workflow. What I’ve found is — I really don’t like Zapier’s limitations. For Zapier, you’ll need to pay around 25€/month just to get hit with a limit of 750 tasks/month. Not on my watch. So long, Zapier!

The next one is Make.com, an incredible piece of automation software. More flexibility than Zapier, more integrations, more automation possibilities, and it’s really non-dev-friendly. This is a big selling point (for me especially) since I’m nowhere near being able to write good code. Still, there’s a limit on how many operations you can complete each month — and that’s something I view as a big minus.

Enter n8n, an automation platform with virtually endless capabilities and no limits. With n8n, you can connect and automate ANYTHING. A bold claim, but so far every case I wanted to prove with n8n has worked wonderfully. It’s a bit more technical than the previous two, but I’ve created a small AI helper to help you navigate the tool more easily (we’ll get to that later).

Whichever of these 3 you go for, the principles are pretty much the same. My honest recommendation is that you use n8n due to its sheer power and the ability to automate anything and improve your workflows with AI agents. And there’s an option for self-hosting — that’s always cool.

Now, let’s talk about the key thing here. How to actually do this stuff?

How do we automate repetitive activities?

I’ve prepared for you 3 concrete use cases, which I have implemented already to free up some time to work on high-impact tasks — and to test the technology. :)

I’ll show you how I created these automations and what it takes to build one for your own needs.

QUICK NOTE: Click here to access a Notion page where I've added a few automations #1 and #3 as ready-to-use templates. Just download the JSON file, open n8n, import the JSON file, and you're all set!

Let’s start.

Automation #1: Topic research and content generation

Here’s one issue we had. We have our own mobility media platform, Charging Stack. There, we publish news from the mobility industry every day, we share our interviews we conduct with mobility industry leaders, and we work on a weekly newsletter.

We had to face it — there’s just a few of us and publishing 5 news stories (that are actually relevant to the date and impactful and well-written) seemed a bit overwhelming. And it happens every day — every day we aim to publish 5 new ones.

I know there has to be a way to speed this up.

So, I opened up my n8n dashboard and started planning.

Here’s my thought process:

• I need to have 5 well-researched and well-written news articles published every day on our website
• To do this, first I need to research — can I use Perplexity for that?
• How can I use that research data?
• Can I hook up an AI agent that can read that research and write news stories?
• After it’s written, I’ll need to write metadata and prepare the article in HTML
• Can I connect it to Wordpress so when it gets pushed to “Draft”, I get notified via email?

After some thinking, the solution was clear.

Below, you’ll see how it looks all connected and fired up.

Once this workflow is done, it generates 5 new article drafts in my Wordpress account. Of course, I had to go and double-check the sources to see if it’s really well-written and if it’s relevant to the news we’re covering. And it all checks out.

This was a simple automation, but check this — it has 2 AI agents and 1 AI model node — that’s like having 2 employees taking their time to help you solve your task.

Automating the creative work is almost never the right answer, but successully implementing this made me think about other possible use cases.

Automation #2: Social media publishing

Now, one other thing that was taking too much time was social media. For each article we publish, we aim to publish it on X and LinkedIn. This means we need to have unique captions for both platforms and they need to be published throughout the day.

Yes, writing this on my own would take only a few minutes a day, maybe 10-15. That’s an hour/week. Now I have one hour more!

Here’s my thought process:

• I want to automate my social media posting so whenever I publish a post, it gets published on our socials
• There should be an AI agent preparing social media captions based on the article content
• There should be a small time gap between when the article is published on our website and on our socials
• I want to get notified when the posts are shared on our socials

Below, you’ll see how it looks all connected and fired up.

This means that:

• Every day, Perplexity conducts deep research
• AI agents handle all the writing
• The articles get sent to “Draft” where they require human editing
• Once a human edits and publishes, everything is prepared for social media and reshared

This is all done automatically! Except the human editing part — this is something I wouldn’t take out of the equation just yet.

Automation #3: Weekly reporting

Since we’re an agency, we have a ton of projects to handle on a weekly basis — and many of the clients have been with us for years. And every week, we want to know what’s the current state of these projects — what are the top pages, which queries are crushing it, where do the users drop off, etc.

Let’s imagine we have 20 projects we are working on. For each project, we want to create a weekly report. And that takes time. Now, that’s fully automated and it’s freed up our project managers’ time by several hours weekly.

Here’s my thought process:

• I want to fetch data from Google Analytics 4 and Google Search Console to analyze
• I want to compare this week’s data with the previous week’s data for both platforms
• It would be great if there was an AI agent that can do the analysis, point out anomalies, recommend next steps, and pinpoint some of the greatest wins/losses of the week for each project
• This report needs to be written in a Google Document and that document needs to be viewable by all people from my company
• I want this report sent as an attachment via Gmail to our stakeholders

Below, you’ll see how it looks all connected and fired up.

This means that:

• Every week, we’ll fetch the data from GA4 and GSC
• We’ll generate weekly reports for both GA4 and GSC
• We merge those reports and use a Code node to make the report easy-to-understand for the AI agent
• The AI agent takes the merged report and writes a detailed analysis
• A Google Doc is created and it’s populated with the weekly report
• Our stakeholders get notified via Gmail

This automation is the most complex out of 3 and it brings incredible value to our team.

But inside these automations, there’s something incredibly powerful — AI agents.

Why are AI agents that important?

One thing that made all the difference for us when choosing the automation platform — the AI agents.

So, in n8n, you can use AI agents in 2 main ways:

1. AI agents can generate an output based on the instructions
2. AI agents can execute tasks, following the instructions

Below, you can see how an AI agent node looks like. It allows you to:

• Choose a chat model — from locally-hosted models to almost any AI model you can connect with using an API key. You even get some free OpenAI credits once you set up your account.
• Memory — you can set up memory so your AI agent can remember and reference its past interactions. The wild thing is - this memory can be shared with more AI agents. This means that all of them can share memory! Incredible.
• Tool — you can choose which tool your AI agent should use. You can write your own tool in form of a Code node (Javascript or Python), or you can choose one from the long list of available ones. Or even use one of your entire workflows as a tool!

The existence of AI agents in these workflows makes the entire difference.

Why this matters so much?

In just a few years, we’ve gone from automations that worked as simple connections between a few tools to multi-agent workflows that can operate on a much more independent level.

Now, we can build systems that operate with much less micromanagement. And AI agents help us reach the next level, especially if we design them with clear roles and tools.

We have now reached a point where:

• One agent can do deep research
• Another can write, revise, and format content
• A third can run checks or trigger decisions based on your logic

All inside a single n8n workflow. If that sounds like having a small team of autonomous interns quietly getting things done in the background — that’s the point.

The best part is that you control everything. There’s no vendor lock-in, no per-task pricing, and no black-box automations.

If you can imagine it, you can most likely automate it. And we’d like to help you do that.

What lessons are there to learn?

In the end, this doesn’t mean replacing people. Our goal is to remove the stuff that slows them down.

After building dozens of these systems, here’s what I’d tell anyone trying to get started:

1. Don’t always automate creative work — this is the stuff that makes us happy just to work on it.
2. Before you start working on a workflow, write out the entire workflow on a whiteboard or paper — it needs to make sense.
3. Things will break — so add alerts, add error workflows, fallbacks, etc.
4. Always track results — if your workflow saves you time, how much? What’s the ROI?

Automation works best when it makes the team feel lighter, not replaced. Want help getting there?

If your team is buried in repetitive tasks, you don’t need more people. You need better systems.

Want help setting them up? Reach out and we’ll build something that perfectly fits your processes.