5. Agents vs Consumer Apps - Basic
Note: The video covers material not in the guide below — please watch in full.
Training Guide
You've used ChatGPT. Maybe Claude.ai. Maybe Gemini. You typed a question, got an answer, copied it, pasted it somewhere else. That's a consumer app.
What you're about to learn is fundamentally different.
An agent platform doesn't just answer your questions — it does the work. It reads your files, writes your documents, connects to your tools, and executes tasks. You don't copy-paste from it. You tell it what you need, and it builds it.
Think of it this way: a consumer app is like texting a really smart friend. You ask, they reply, you do something with the reply. An agent platform is like hiring an assistant who sits at your desk, has access to your computer, and does the thing you would've done yourself — but faster.
(This is the most important distinction in the entire course — everything else builds on it)
What Consumer Apps Actually Do
To be fair — consumer apps aren't standing still. ChatGPT has Custom GPTs & ability to connect to other apps. Claude.ai has projects and can also connect to external tools. They're getting more capable every month.
But the core workflow hasn't changed. You go to a browser tab. You type a question. You get text back. And then:
- You copy that text
- You paste it into a doc, an email, a spreadsheet
- You format it, fix it, organize it
- You do this over and over
The AI did the thinking. You still did the work.
And there are things you just can't do in that setup:
- You can't point it at five files across three folders and say "cross-reference these" (that's called file system access — you'll see this in action soon)
- You can't see how much memory the AI has left or what it's actually working with (that's called token management — we'll get into it)
- You can't give it a set of instructions it remembers across every conversation (those are skills and SOPs — they change everything)
- You can't tell it to go research ten things at once while you keep working (those are sub-agents — yes, it can clone itself)
A consumer app lives in its own world. It doesn't know what's on your computer. Even when it connects to something, it's one thing at a time, inside its own sandbox — not yours.
What Agent Platforms Do
Now here's what happened when you set up Claude Code in the setup training. You installed something that doesn't live in a browser tab. It lives on your machine. Inside your file system. Next to your actual work.
That one difference changes everything.
You don't go to the AI — the AI comes to you. It's already sitting in your workspace, looking at your folders, ready to work with whatever you're working on. You don't copy-paste into it. You don't copy-paste out of it. You just tell it what you need, and it does the work where the work lives.
The things you just heard that consumer apps can't do? File access, memory management, sub-agents, reusable skills — those aren't add-ons. They're what happens when the AI is operating in your world instead of its own.
(So what does that actually give you? Here are the specific abilities)
Agent Platform Specific Abilities
Creating files is just the start. Here are the 8 things you now have access to that don't exist — or barely exist — in a consumer app.
1. Permission Modes
You control how much freedom the AI has. Want it to ask before every edit? Done. Want it to create a plan for your approval before executing? Done. Want it to just go — full autonomy, no hand-holding? That's an option too. You set the level of trust based on the task.
2. Queue Up Messages
While the AI is working on your first request, you can already type your next one. It queues up. In a consumer app, you wait for the response, read it, then type again. Here, you can stack instructions and keep moving.
3. File System Access
This is the big one. The AI can read, write, and edit files directly on your computer. Not in a temporary sandbox you download from — in your actual folders, alongside your actual work. You point it at a file, it opens it. You tell it to create something, it's saved where you said.
4. Todo Lists
Tell the AI: "Make a comprehensive to-do list for yourself." It will. It breaks down your request into steps, tracks its own progress, and checks things off as it goes. A consumer app can't self-organize — it just responds to one message at a time.
5. Token Management
Every AI has a memory limit — a budget of how much it can hold in its head at once. Consumer apps hide this from you. You just hit a wall and start over. An agent platform shows you exactly how much memory you've used, and gives you tools to manage it — compacting conversations, saving context, handing off to a fresh agent when you need more room.
6. Sub-Agents
The AI can spin up copies of itself. Need to research 10 competitors? Instead of doing them one by one, it spawns 10 sub-agents that each handle one — in parallel. Each gets its own fresh memory. Think of it like a manager delegating to a team.
7. Skills & SOPs
You can give the AI persistent instructions — reusable playbooks that tell it how to do specific tasks your way. Get a great output once, turn it into a skill, and now the AI can repeat that process every time. It's like uploading expertise to its brain.
8. MCPs — Connecting to External Tools
MCPs let the AI talk directly to other platforms — your project management tool, your CRM, your calendar. Instead of opening five tabs and manually pulling data, you ask the AI and it pulls it for you. Consumer apps are starting to get basic tool connections, but the depth and flexibility here is on another level.
(Now let's see the difference in action)
Side by Side: The Same Task, Two Different Worlds
Here's a real example — this course you're watching right now.
To build this training, I needed a teaching methodology file, a course index, a master outline, and raw source material from previous trainings. That's four documents across different folders.
In ChatGPT (consumer app):
- Open the methodology file, select all, copy
- Paste it into ChatGPT
- Go back, open the master outline, select all, copy
- Paste that in too
- Now explain what you want
- Read the response
- Copy the output
- Open a new file, paste it in
- Realize you need to reference the course index — go back, copy it, paste it into a new message
- Repeat for every edit, every revision, every cross-reference
In Claude Code (agent platform):
- "Read the teaching methodology, the course index, and the master outline. Now write the next training based on the outline bullets and save it to the S2 folder."
- It reads all three files, writes the training, and saves it. If I want an edit — "update the methodology to include the new connector format" — it opens the file, makes the change, done.
That's the difference. One workflow has you bouncing between tabs, copying and pasting context that the AI forgets between messages. The other has the AI sitting in your project, seeing everything, working across files the way you would.
(So why did we choose this particular agent platform?)
Why Claude Code
There are several agent platforms out there — Claude Code isn't the only one. Anthropic also has Co-Work. OpenAI has Codex. Google is building their own. The landscape is moving fast and we'll break down all of them later.
We're using Claude Code because at the time of making this training - it's the one built for b2b and focused on that use case. Anthropic's model also are the best for "knowledge work" where OpenAI and Google's agent models are less conversational and more coding / engineering focused.
But again, this is changing very rapidly - the key to know is: when you learn the skill of learning with Agents in general, the skill set transfers. There's some nuance and differences but for the most part, it's a transferrable skill.
(We'll go deeper into the full landscape — what else is out there, how they compare, and why this matters — in a later training)
The Thing Most People Don't Know
Here's the key unlock: a lot of people don't even know agent platforms exist. They're still in the copy-paste era — using ChatGPT or Claude.ai in a browser, manually moving text around, doing all the formatting and organizing themselves.
They think that's what AI is. It's not. That's just the consumer version.
You're learning the version that most people haven't discovered yet. The tools that actually do the work — not just suggest it.
That gap is real. And the longer it takes for people to cross it, the bigger your advantage.
(Over the next few trainings, I'm going to show you each of these abilities one by one. Let's start with the controls)
Comment in Slack
Post your answer in your onboarding channel.
What was your biggest takeaway(s) from this training?