01What Is the First Automation to Build in Agent OS?

The first automation to build in Agent OS is the one that saves you the most time. You find it by looking at your daily tasks. This episode shows a simple way to pick and build that automation.

The process has three steps. First, watch where you spend your time. Second, build that task into your Agent OS.

Third, test it and fix problems. This repeats every week, so your system gets better over time.

02Step 1: Find the Task You Repeat Daily

The first step is to find what you do over and over. Julian says this is different for everyone. What you spend time on is not what someone else spends time on.

So you need to look at your own work.

If you are not sure, track your time for one week. Write down what you do every hour. Look for tasks that repeat daily.

That is your best target for automation.

For example, a member named Mazzy asked about automating ads. She knew she spent time on Meta ads reporting. So that was her first automation.

Another member, Sapan, just installed Agent OS and wanted to know what to do next. Julian says to use the same idea.

03Step 2: Build the Task Into Your Agent OS

Once you know the task, you build it into your Agent OS. Julian explains how he did this for SEO. He built a keyword research section because he spent a lot of time on that.

He also built a way to generate content by typing a target keyword, adding a case study, and publishing it to his website.

To build an automation, you use an agent harness like Codex, DeepSeek, or Code. You describe the workflow you want. Then you ask the agent to start building tests.

Today, Julian asked Codex to add a section that generates new ads including creatives and copy. He wanted it to look at existing ads and make new ones that are just as powerful. It took about six minutes to build that workflow.

Now he has an 'Ad Studio' in his Agent OS.

04Step 3: Test Your New Automation

After building, you must test the automation. Julian shows how he tested the ad generator. He entered 'SEO' as the product, 'website owners' as the audience, and gave creative direction.

He asked for a different hook and simpler language.

The tool gave an error. That is normal. Most automations do not work the first time.

When you see an error, go back to the agent that built it. Give the error message or a screenshot. Ask the agent to fix it and test it itself.

This saves you time.

In the example, Julian gave the error screenshot to Codex. Codex then started debugging the automation. This back-and-forth is part of the process.

Over time, the automation improves.

05Why You Should Build Automations Weekly

Julian explains that this process never stops. Every week you loop through the three steps. Each week you spend time on something new.

That is a new chance to build another automation.

For example, Julian started with the memory system for Agent OS. Then he built the SEO system. After that, he built the video agent, the image generator, and the Gemini notebook connection.

He keeps adding new automations.

By focusing on one automation per week, you avoid feeling overwhelmed. You build your system step by step. Over time, your Agent OS becomes more powerful and saves you more time.

06Real Examples From the Episode

This episode shows two real examples. One is Sapan, who installed Agent OS and Hermes agent. Sapan added an OpenRoute API and got Hermes responding.

Sapan asked what to do next. Julian says to identify the next time-consuming task and automate it.

Another example is Mazzy. She asked to automate ads with Agent OS. Julian and his team built that integration for her.

They built the ads reporting and ad generation feature directly into the system.

Julian also mentions his own SEO automation. He built keyword research and content generation. These examples show that you can automate many different tasks with Agent OS.

07Tips, Limits, and Mistakes to Avoid

Here are key tips from the episode:

  • Start with one task per week. Do not try to automate everything at once.
  • Track your time if you do not know what to automate. One week of recording hourly tasks will show you.
  • Expect errors when testing. Most new automations do not work first time.
  • Give clear feedback to the agent when fixing errors. Screenshots help.
  • Ask the agent to test itself so you do not waste time testing broken code.
  • Loop every week. Your system improves if you keep building.

Mistakes to avoid include rushing through steps and ignoring errors. Also, do not pick a side topic that does not save you time. Stick to tasks that repeat daily.

08Who This Is For and Why It Matters

This is for anyone who uses an agentic operating system and wants to make it useful. Whether you are new like Sapan or more advanced, the same process works. It matters because it turns a general tool into a personal automation machine.

By building the first automation that matches your work, you start saving time right away. Over weeks, you build a system that handles many of your daily tasks. This lets you focus on bigger work instead of repeating the same steps.

Julian also offers help in the AI Profit Boardroom. Members can ask questions, get video tutorials, and receive step-by-step guides. He even builds automations for members, like he did for Mazzy.