I recently spoke at CloudNexus, an AWS Cloud Club QAU event focused on cloud technologies and AI. It was a short talk, nothing fancy, mainly around Agentic AI and how students are currently using AI tools.
What surprised me wasn’t the questions. It was how confident students were that they were “using AI correctly”. Most of them weren’t.
“I asked them to generate code from ChatGPT”
During discussions, a lot of students openly said they use ChatGPT mainly for code generation. Not for understanding. Not for learning concepts. Just for getting code quickly.
And honestly, I would not blame them for this. It’s very fast. It feels somehow productive. But this is where problems start showing up later during debugging, real projects, or even basic system design conversations or in interviews.
The issue isn’t AI: it’s how we’re using it
One thing I tried to explain during the talk was simple:
Not every AI tool is meant to do the same job.ChatGPT is great. I use it too. But it shines more when you use it for:
- understanding concepts
- breaking down ideas
- writing or structuring content
- asking “why” questions
When you try to use it as your main coding engine, you miss context and context matters a lot in real-world development.
Code-focused tools exist for a reason
We talked a bit about tools like Cursor, Amazon Q etc. Thesetools work inside your codebase. They understand files, references, and structure.That’s why they feel more useful for development work.
Not because they are “smarter AI”, but because they are built for developers.
Where Amazon Q fits in
Since this was an AWS-focused event, I also talked about Amazon Q. What I personally find interesting about Amazon Q is that it doesn’t live outside your workflow. It helps inside AWS environments, where cloud engineers actually spend time. It’s less about asking random questions and more about:
- cloud-aware guidance
- best practices
- security-conscious suggestions
That difference matters when you move beyond tutorials.
Explaining Agentic AI in simple terms
To keep things simple, I explained AI in three phases:
- Chatbots: you ask, it replies
- Copilots: it understands your context
- Agents: you give a goal, it figures out the steps Agentic AI is less about prompts and more about delegation.That idea really clicked with students.
Why the cloud becomes unavoidable here
One thing I emphasized was that agent-based systems don’t work in isolation.They need infrastructure, scale and security.
This is where platforms like AWS Bedrock start making sense, not as buzzwords only, but as enablers for building AI systems that actually run in the real world.
What I really wanted students to take away
If I would summarize the talk in one line, it would be this: AI should help you grow, not replace your thinking.
- Use ChatGPT to learn.
- Use coding copilots to build.
- Use cloud AI to scale. And use agents when you’re ready to automate outcomes.
Final thought
Speaking at CloudNexus reminded me that students don’t need more tools.They need better guidance on how to use what already exists. There is a quote I read online saying:
AI isn’t going away.But engineers who understand how and why they use AI will always stand out.
here's the linkedin post for my recent talk:
#cloudnexus #cloudnative #agenticai #awsome #aws | Muhammad Abdullah Haroon | 13 comments
𝐆𝐫𝐚𝐭𝐞𝐟𝐮𝐥 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐨𝐩𝐩𝐨𝐫𝐭𝐮𝐧𝐢𝐭𝐲 𝐭𝐨 𝐣𝐨𝐢𝐧 𝐂𝐥𝐨𝐮𝐝 𝐍𝐞𝐱𝐮𝐬 𝐚𝐭 Daftarkhwan vanguard 𝐚𝐬 𝐚 𝐰𝐢𝐥𝐝 𝐜𝐚𝐫𝐝 𝐞𝐧𝐭𝐫𝐲 𝐬𝐩𝐞𝐚𝐤𝐞𝐫 🙃 I had the chance to represent Cloud Native Islamabad and its ecosystem and share thoughts on emerging technologies and opportunities in Agentic AI and Cloud, with a special focus on how the Cloud Native Computing Foundation (CNCF) ecosystem (including CNI) is shaping the future of cloud-native engineering. Huge shoutout to Ali Mumtaz for organizing such a well-executed and impactful event Also, appreciation for Sharoon Amin Akhtar for sharing valuable insights on serverless cloud technologies. The panel discussion was equally engaging, featuring DevOps Molvi (Faisal Rehman), Hamza Nasir 🚀 Tameem Ud Din and Yahya Qureshi , great perspectives and practical discussions around DevOps and modern AI technologies. Lastly, thanks to my The Computer Science Society (CSS) AWS Cloud Club IIUI team for accompanying me Junaid Hanif Shah Murad Ali Shehzad Nazir Sardar Huzaifa Muhammad Sayyam Abbasi Shahzaib Ashraf Uzair Akhtar Gondal praying and best wishes to my mentor Saim Safder who didn't join us because of an incident.. Looking forward to more such conversations and collaborations #CloudNexus #CloudNative #AgenticAI #awsome #aws Amazon Web Services (AWS) AWS Developers Cloud Native Computing Foundation (CNCF) Audra Montenegro Lisa Bagley, CPACC | 13 comments on LinkedIn
linkedin.com
Top comments (9)
But the Real Question is What are they learning Using Even Agentic AI? That just makes them More lazier and Dependent not that using AI shouldn't be promoted I am not saying that but there is a constant Decline of Learning from AI or Using it to Increase your Knowledge and Use it for generating code you Understand
Because as You said It gets serious when the Debugging starts when Errors come knocking on the door thats where the developer comes in and AI Works as a side Assistant
Soo I hope they learn something from Agentic AI not become fully Dependent on it as they are dependent on GPT and other ai models
Agree 100%. This is the biggest problems of this generation. If we look at it from another angle, those who ethically criticize AI are not totally wrong. Yes, we are becoming totally dependent on it. We are now using AI even for messaging and replying to others, which is removing emotions and authenticity from communication.
The same issue appears in development side as well. Someone who doesn’t understand DSA concepts or can not even read the code can build a website using AI. I shared an example with them about a famous tweet where a person claimed that he built a fully functional website with just $20 using AI. A few days later, the same person admitted the project had serious loopholes, and attackers easily exploited it.
My point to them was simple that use AI to handle repetitive or time consuming tasks and only for productivity, but if you’re generating code with AI, you must be capable of reading, understanding, and debugging it yourself. AI should assist your learning, should not replace it.
attended the session really helpful in my machine learning career..... I always thought AWS is all about cloud only but didn't heard about amazon q or bedrock or kiro........ really learn alot from you sir...............
Thanks zaheer
Thanks for sharing I also used Amazon Q, i think its a very powerfull
Thanks ahmed
👏 👏
😊❤️👏👏
awsome