How To Learn Creating AI Agents (Without Writing Code)

How to learn creating AI agents is much easier when you treat it like hiring, because you write the job description first and choose the software second.

I'm Julian Goldie, and I run a 7-figure SEO agency where AI now handles a large share of the internal work.

I've learned that the people who get value from agents are rarely the most technical ones.

They're the ones who can describe one job clearly.

So here's the path I'd give any business owner, agency owner or creator.

You pick one repeated job in your week that has a clear finished result.
You write a one-page agent brief, which is the job description for the agent.
You choose a brain, and you start with a free model or a subscription you already pay for.
You create the agent in a free tool and keep it on "ask for approval".
You test it on five real pieces of your own work.
You break it on purpose, fix it and only then let it run on a schedule.

I'll show you the whole thing with one agent, using the real creation screen from my StarNet video.

How To Learn Creating AI Agents: The Hiring Analogy

An AI agent is a model that has a job, instructions, tools and something that starts it.

That sounds technical, but you've already done the same thinking every time you've hired or briefed someone.

When you hire a person	When you create an agent	What you decide
You write the role.	You define the job.	You name the one outcome it delivers.
You choose the candidate.	You choose the brain.	You pick the model that does the thinking.
You give them a handbook.	You write the instructions and persona.	You set the rules and the tone of voice.
You hand over logins.	You pick the tools and permissions.	You decide what it can use and what it can touch.
You expect them to remember clients.	You set up memory.	You decide what carries over between sessions.
You agree when they work.	You choose the trigger.	You decide whether chat, a schedule or a message starts it.
You sign off their first tasks.	You set approvals and guardrails.	You decide when it must ask you first.
You run a trial period.	You test it.	You run five real tasks and judge them.
You give feedback.	You improve it.	You change one thing at a time.

Nobody sensible hires first and writes the role afterwards.

Most people do exactly that with agents, which is why their agents disappoint them.

This guide is about designing and creating one agent properly.

If you'd like a ladder of projects to work through afterwards, read how to learn building AI agents next.

Start With The Job, Because The Job Is The Business Case

The job is the only part of an agent that your business actually cares about.

A good agent job has three features.

It happens often, such as every day or every week.
It has a finished result that you can check in a minute.
It's safe if the first few attempts are wrong.

"Grow my business" fails all three tests.

"Turn each call transcript I paste in into clean notes with action points" passes all three.

I teach one agent doing one job, because a small job is easy to check and easy to fix.

I won't promise you any income from an agent, because I can't see your business.

What I can tell you is that a clear job is the difference between an agent you use daily and one you abandon.

🔥 Want my exact How To Learn Creating AI Agents setup, step by step? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom

Three Example Jobs: Owner, Agency And Creator

Here are three small jobs that suit a first agent.

Who you are	A good first job	What finished looks like	The trigger I'd start with
A business owner	It reads my task list and writes a short to-do recap.	I get one short message each morning.	Chat first, then a 6am schedule
An agency owner	It reads the search queries I paste in and names three to write about.	I get three keywords with one line of reasoning each.	Chat
A creator	It turns a video transcript into notes and action points.	I get one clean page of notes per video.	Chat

The business owner's job is a real one from my own Hermes guides, and I'll show you the command for it further down.

The agency job is the one I'd create first myself, because SEO is what I told my own StarNet agent I wanted help with.

The Agent Brief: Your Fill-In Job Description

You fill this in before you open any tool.

AGENT BRIEF
Agent name:
The one job (one sentence):
Finished looks like:
Who checks the work, and how often:
Brain (model):
Reasoning level:
Personality:
Instructions (5-10 plain sentences):
It must never:
Tools it needs (max 2):
What it may read:
What it may change:
What it should remember:
What starts it (chat, schedule or message):
Approval level:
Five real test tasks:

The line that business owners skip is "it must never".

You should write down at least two things, such as never sending a message to a client and never inventing a number.

Those two lines are your first guardrails.

Creating One Agent End To End: A Real Walkthrough

I'll use the agency job from the table, and I'll create it in StarNet.

StarNet is a free, open-source agent harness that shows your agents inside a pixel-art space station.

I'm using it here because every design decision in the brief shows up as a button on the creation screen.

The brief, filled in

Brief line	The answer for this agent
Agent name	Scout, which is only my example name
The one job	It reads the search queries I paste in and names the three I should write about next.
Finished looks like	I get three keywords, each with one sentence of reasoning.
Brain	I use Grok through the account I already pay for.
Reasoning level	I choose medium.
Personality	I choose composed.
It must never	It never invents a search number, and it never publishes anything.
Tools	It gets web search only.
What it may change	It may change nothing.
What starts it	I start it from chat.
Approval level	It asks for approval.

The clicks, in order

You give the agent a name.
You see its appearance, and you can change it.
You pick a personality, and in my video I talked through warm, dry, composed and unhinged.
You choose the approval level, which is "ask for approval" or "full power".
You connect a brain by signing in to an account or by pasting an API key.
You pick the model and the reasoning level, which runs from low through medium and high up to max.
You create the agent, and a "waking up" animation plays.

I signed in with Grok in one quick step, so there was no new bill to set up.

I chose Grok 4.6 over Grok 4.7, because I actually prefer Grok 4.6 for that kind of conversation.

My agent's first words were "I'm awake. Let's give the station a direction."

It then asked me, "What made you want to set up an AI agent?"

I told it SEO and gave it aiprofitboardroom.com.

The settings you finish afterwards

You click the agent to open its detail panel.

The panel shows the model, the personality, the instructions, the access permissions and the approval prompts.

You paste the instructions from your brief into the instructions field, which holds the agent's purpose.

You then set the access permissions, which cover things such as the web and your local files.

You can allow or remove each one.

The panel also has tabs called Brief, Growth, History and Memories.

History shows what the agent has done, and Memories shows what it has remembered.

Those two tabs are where a business owner should look once a week.

My honest view of StarNet

I said in the video that I think StarNet is fun to play with, and that I don't think there's a massive amount of practical use cases yet.

I still think that.

There's also a slight delay when you send a message.

I recommend it for learning, because you can see every decision you're making.

My StarNet AI agent guide shows how a business might use the whole station.

🔥 Want agent briefs you can copy for your own business? Inside the AI Profit Boardroom, I share the agent setups I run, my installable Agent OS and courses on Hermes Agent, Grok Bot, DeepSeek Harness and Claude Opus 5.5. You'll be learning alongside 3,400+ members. → Get access here

Choosing The Brain Without Wasting Money

The brain is where beginners overspend.

You don't need the most expensive model to learn agent creation.

You have three sensible options, and I'd try them in this order.

Option	What it means	Why a beginner should care
A free model	You use a model that costs nothing to run.	You can make mistakes without watching a meter.
A subscription you already own	You sign in with an account you pay for.	There's no new bill to approve.
Pay per use	You paste an API key and pay for what the agent uses.	You get the widest choice, and you need to watch the cost.

In Hermes Agent I keep one profile per model.

A profile is a separate folder with its own settings, memory and plugins, so each brain gets a clean agent.

That makes it easy to run the same job on two brains and compare the results.

You create a profile with hermes profile create followed by a name, and you give it a model with hermes -p followed by the name and the word setup.

My guide to the best Hermes Agent model lists the brains I use and what each one costs.

Design Choices And What To Pick As A Beginner

Design choice	What I'd pick first	How it protects your business
The job	One outcome	You can check the work in a minute.
The brain	Free or already paid for	You learn without a surprise bill.
The reasoning level	Medium	You get sensible answers at a sensible speed.
The personality	Composed	The tone stays professional.
The instructions	Five to ten full sentences	The agent has clear rules to follow.
The tools	One or two	There's less for the agent to get wrong.
The permissions	Read only	It can't change or delete anything.
The memory	The built-in memory	You can read exactly what it has stored.
The trigger	Chat	You see every result before it matters.
The approval level	Ask for approval	Nothing happens without your yes.

My rule for permissions is that an agent gets only the access its job needs.

My rule for approvals is that every new agent starts on ask for approval, and I loosen the leash once I trust it.

That's how I'd onboard a new team member too.

Five Agents Worth Creating First

These five are small enough to test and useful enough to keep.

A daily recap agent reads your task list and sends a short to-do summary.
A keyword scout reads your search queries and suggests what to write about next.
A draft-reply agent writes replies to messages you paste in, and you send them yourself.
A notes agent turns a call or video transcript into notes and action points.
A research agent answers one question with a short summary and its sources.

None of these touch a customer directly, and that's deliberate.

Your first agents should work for you, with you as the check, before any agent works in front of a client.

The daily recap is the one I'd schedule first once it's passed its tests.

This is the Hermes command from my own guide.

hermes cron create "0 6 * * *" "Read my task list and send me a short to-do recap for today" --name "Daily 6AM To-Do Recap" --deliver telegram

You can read the brief straight out of that line.

The schedule is the trigger, the prompt is the job and Telegram is where the result is delivered.

My guide to the best Hermes Agent use cases has more jobs like this.

If you'd like a second pair of eyes on your own business, you can book a free SEO strategy session and we'll map out your plan together. If you're already serious about SEO and want to work with other experienced operators, have a look at the SEO Elite Circle.

A Four-Week Practice Plan: Create, Break, Fix

You'll learn creating agents faster with a simple plan than with another course.

Week	What you do	What you should have at the end
Week 1: Create	You write one brief and create one agent on chat.	You have one agent and five test results.
Week 2: Break	You try to make it fail on purpose.	You have a list of its weak spots.
Week 3: Fix	You change one thing at a time and re-run the tests.	You have an agent you trust on that job.
Week 4: Repeat	You schedule the first agent and brief a second one.	You have one agent running and one in testing.

How to break an agent on purpose

You give it a vague request and watch whether it asks a question or guesses.
You ask for something outside its job and watch whether it says no.
You paste in messy or incomplete input and check the result.
You swap the brain for a cheaper one and run the same five tasks.
You remove a permission and see what it does when it can't reach something.

Every failure tells you which line of the brief to rewrite.

A vague answer usually means the job line is vague.

A made-up fact usually means the "it must never" line is missing.

An agent that does too much usually has too many tools.

How to fix it

You change one line, one tool or one setting.

You then run the same five tasks again and compare the results.

If the agent improves, you keep the change.

If it doesn't, you undo the change and try a different one.

What A First Agent Won't Do For Your Business

I want to be straight with you about the limits.

A first agent won't replace a member of staff.

It won't fix a job that you can't describe clearly yourself.

It will make mistakes, which is why it asks for approval.

It also needs you to read its work for the first few weeks.

The payoff is that you learn a skill you can reuse, because every agent after the first one uses the same brief.

🔥 Want feedback on your first agent brief? The AI Profit Boardroom runs four coaching calls a week, so you can bring the brief for your business and get it checked by me and the team before you create anything. → Join the AI Profit Boardroom

Also On Our Network

FAQ: How To Learn Creating AI Agents

How do I learn creating AI agents if I'm not technical?

You start with the job and not the software.

You write a one-page brief that names one outcome, the rules, the tools and the approval level.

You then create the agent in a tool that uses clicks, such as StarNet, and test it on your own work.

What should my first business agent do?

Your first agent should do one repeated job with a result you can check in a minute.

Good examples are a daily to-do recap, transcript notes or a keyword shortlist.

It should work for you and not in front of customers.

How much does it cost to create an AI agent?

It can cost nothing to start.

StarNet and Hermes Agent are both free and open source, and you can use a free model or a subscription you already pay for.

You only pay per use if you choose to paste in a paid API key.

Is it safe to give an agent access to my business tools?

It's safe when you limit the access to what the job needs.

I start every agent on read-only permissions and on ask for approval.

I only widen the access after it has passed its tests.

What is an agent brief?

An agent brief is a one-page job description for an AI agent.

It lists the job, the brain, the instructions, the tools, the permissions, the memory, the trigger, the approval level and five test tasks.

You fill it in before you create anything.

How long does it take to learn creating AI agents?

There's no fixed timeline, because it depends on how often you practise.

My four-week plan is one week to create, one to break, one to fix and one to repeat.

You can create the first agent in a single sitting.

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

About Julian

I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.

I help business owners scale with AI agents, automation and SEO.

I built Goldie Agency, a 7-figure SEO and link building agency with 70+ employees.
I've grown a YouTube channel to 400,000+ subscribers.
I wrote the book "Link Building Mastery".
I built the Agent Operating System, which runs a crew of AI agents under one shared memory.

→ Get my best AI training inside the AI Profit Boardroom

Write the job description, create one agent and practise on your own work, because that's how to learn creating AI agents.