When an AI Coding Assistant Taught Me a Bigger Lesson About the Future of Software Development

During a recent training session, I had one of those rare moments where the teacher becomes the student.

I was conducting a hands-on session on using an AI coding assistant with a group of nearly 1000 learners I've trained so far. As usual, we ended with a Q&A. That's when one of the attendees asked a simple, practical question:

"I work a lot with Excel files in .xlsx format. Can this AI coding assistant directly read .xlsx files?"

I hadn't tried that specific scenario before, so I was honest: "I'm not sure. Let's test it live."

And that's where things got interesting.

The Live Demo That Surprised Even Me

To explore the question, I quickly created an .xlsx file with some dummy data and uploaded it to the AI coding assistant.

Its response was:

• It couldn't read the .xlsx file directly.
• It offered me three options:
  • Copy-paste the content as plain text
  • Save the .xlsx file as a .csv and upload that
  • Let it write a script to read the .xlsx file programmatically

Naturally, I chose the third option.

The assistant then:

1. Generated a script to read the .xlsx file (using a suitable programming language and library).
2. Read the file through that script and presented the data back to me.

Satya Nadella's Point Clicking in Real Time

This immediately reminded me of something Satya Nadella mentioned in an earlier interview.

He talked about how most typical CRUD (Create, Read, Update, Delete)-based SaaS applications are essentially:

• A UI layer on top of
• A database or some data source

His point was that AI agents will increasingly:

• Read data from multiple sources
• Interact with them in different formats
• And, when they can't directly interact with a source, they will generate the necessary script, interface or adapter themselves.

What sounded theoretical in that interview suddenly became real in my demo.

I saw it live:

• The AI couldn't directly read .xlsx
• So it offered to write a script compatible with that format
• Then used that script to get the job done

Why This Should Concern (and Excite) Developers

Many of us have built our careers around:

• Designing UIs
• Connecting to a database
• Performing CRUD operations
• Showing results on a screen

If our "value" is limited to:

"I can build a UI + connect it to a database + perform CRUD"

Then we are competing directly with AI agents that can:

• Generate forms
• Generate APIs
• Generate database queries
• Generate scripts
• Or even bypass UIs altogether by talking directly to the data source

That will not be enough in the near future.

AI agents are not just another tool in the stack. They are becoming an active layer between users and systems.

What they did with my .xlsx file is exactly what they'll do with:

• APIs
• Legacy systems
• Different databases
• Files and formats we don't even think about today

So, What Skills Will Really Matter?

This experience reinforced something important for me — and it's what I now tell every learner I train:

If you want to stay relevant, you need to go beyond:

• "I know a framework"
• "I can design screens"
• "I can write CRUD operations"

You need to develop skills like:

In simple terms:

The game is shifting from "Can you code this?" to "Can you orchestrate humans, systems, and AI to solve real problems?"

Learning While Teaching

For me, that session was special.

I went in as the trainer, explaining how to use an AI coding assistant. I walked out with a renewed appreciation of just how powerful these AI agents already are — and how much more powerful they're going to become.

It was a humbling reminder that:

• Every time we teach, we also learn.
• Every demo can expose us to a new capability we hadn't noticed yet.
• And every "small" question from an attendee can trigger a big realisation about the future of our work.

A Message to Developers and Tech Professionals

If your current skill set is:

"I know how to build a UI, connect it to a database, and perform CRUD"

You're standing on shrinking ground.

Start moving towards:

• Understanding systems end-to-end
• Designing workflows where AI is a first-class collaborator
• Building solutions that are AI-native, not just "AI-assisted"

Because the future will belong to those who can:

Use AI not just as a tool, but as a partner in building, reading, transforming, and operating systems.

And sometimes, that future reveals itself not in a conference keynote, but in a simple question about an .xlsx file during a training session.

If you found this useful or thought-provoking, I'd love to hear:

• Have you had a similar "AI surprise" moment in your work?
• What skill are you personally focusing on to stay ahead in this AI-first world?

Let's learn together — even while we teach.