Nowadays, AI has become an essential assistant in coding tasks and is increasingly integrated into many product development services — from code refactoring and test generation to analyzing complex logic. But the real question is: how are developers actually using AI?
At PRIMO Tech-a-Break, we’d like to introduce you to the 5 levels of AI usage — not just using AI, but using it effectively — so you can design and develop your product with maximum efficiency.
But the real question we want to ask in PRIMO Tech-a-Break today is:
What level are developers really at when it comes to using AI?
Because there’s a big difference between just using AI and knowing how to use it well.
Level 1: Asking when you're stuck
You use AI like a faster search engine:
What’s wrong with this syntax?
How do I map an array more concisely?
How do I optimize this SQL query?
At this level, AI helps speed things up when you hit small roadblocks —
but it also comes with the risk of blind trust.
Level 2: Laying the groundwork, then building on it
This works best when you already know what to do but want to skip repetitive setup:
Generate a class or schema
Create basic test cases
Scaffold an API structure
If you’re doing full-stack work, you’ll likely use AI at this level a lot — especially for setting up projects quickly.
Level 3: Using AI as a sparring partner
Here, you’re not asking AI to write for you —
you’re using it to think with you:
Ask it to explain your logic back to you
Compare approach A vs. B
Spot holes in your reasoning
Personally, I use this level often.
It helps me see my own thoughts more clearly, just like when you explain your idea to someone else and suddenly understand it better yourself.
Level 4: Understanding intent, not just commands
Many think this only applies to chatbot developers —
but in reality, every dev needs this skill.
Example:
- A user says, “I want the page to feel cleaner.”
→ They might not want you to remove everything — just make it feel less cluttered.
This skill is crucial in product development,
because what people say ≠ what they actually need.
(Just like I talked about in the last post on intent classification.)
Level 5: Let AI take over where it excels
Today’s tools are getting good enough that you can give AI full tasks:
Refactor code to use async/await
Fix failing tests and explain the changes
Rewrite logic to handle edge cases more robustly
If you set up your dev environment properly and know which tasks to delegate,
AI can take a lot off your plate —
but reviewing and making the final call is still on you.
So… what level should you aim for?
There’s no one-size-fits-all answer.
It depends on your work and your goals.
But here’s what I do believe:
Before you teach AI to understand others, you need to understand the “user” or the context first.
Because a developer’s job isn’t just to write correct code —
It’s to write code that serves real people.
Thanks for reading this far.
If you’re using AI at other levels — or see things differently — feel free to share!
I believe how we use AI says a lot about how we think as developers.
