Something clicked this week that I want to share with anyone building on IBM’s AI stack.
I’ve been working on an AI agent project with IBM Bob — and what started as a simple planning session turned into a fully connected development pipeline much faster than I expected.

The goal of this post is simple: to show how IBM Bob can dramatically improve the way I work with AI notebooks every day, without going into code details.

The Pain Point: Debugging Directly in Studio

Developing and debugging Jupyter notebooks directly inside watsonx.ai Studio can be frustrating.

• The editor is fine for quick experiments, but for real projects I miss a proper IDE experience.
• Navigation and debugging get cumbersome.
• Every small change means re-running cells manually and switching context.

Over time, this slows down development and makes it harder to keep projects clean and well-structured.

Instead of just accepting this, I decided to treat it as an AI problem: I explained the situation to IBM Bob in natural language.

Talking to Bob in Natural Language

I described to Bob:

• How I currently work in watsonx.ai Studio
• What slows me down in my daily notebook workflow
• What my ideal development loop would look like

I did not start with a technical specification. I focused on the problems and the outcomes:

• I wanted to keep using watsonx.ai Studio as my execution environment.
• I wanted GitLab to be my central place for version control and collaboration.
• I wanted my local IDE to be the main place for editing and debugging.

This is where Bob really started to shine.

Bob as a Co-Worker, Not Just a Code Generator

IBM Bob helped me think through the architecture of a better workflow and turned the idea into a concrete development loop.

Bob supported me in:

• Understanding how to interact with watsonx.ai Studio via the Watson Data API
• Designing how notebooks should move between Studio, GitLab and my local machine
• Planning the steps needed for authentication, asset handling and keeping everything in sync

The outcome is a three-way flow that I now use every day:

1. watsonx.ai Studio remains the place where notebooks are executed and run.
2. GitLab is the system of record for versions, history and team collaboration.
3. My local IDE is where I edit, refactor and debug with full tooling support.

Instead of manually downloading and uploading notebooks and copying files around, there is now a smooth, repeatable loop that keeps everything aligned.

A Daily Working Setup – Not Just a Demo

The most important part: this is not a theoretical concept and not a one-off demo.

• It is a working setup that I rely on in my daily work as an IBM Champion and consultant.
• Bob did not replace me as a developer – it became a co-worker that understands my goals.
• It helps me explore the right APIs, structure the workflow, and make the development cycle robust.

For me, this changed the way I think about AI assistance:

• It’s not only about generating snippets.
• It’s about helping design and implement the way we work with AI projects end to end.

The combination of IBM Bob and watsonx.ai Studio turned a painful, manual process into a CI/CD-style development cycle for AI notebooks that feels natural, efficient and reliable.

Why This Is Cool for Anyone Building on watsonx.ai

If you are building on IBM watsonx.ai and still struggle with your development workflow, it is absolutely worth taking a serious look at what Bob can do for you.

You can:

• Describe your problems and desired outcomes in your own words.
• Let Bob help you design the architecture and workflow around watsonx.ai Studio.
• Move from ad-hoc manual steps to a repeatable, automated development loop.

For me, this has been a real game changer. IBM Bob does not just “write some code” – it actively helps shape how I work, and it makes my daily AI engineering on watsonx.ai smoother and more productive.

About the Author

Dr. Michael Straif
Manager AI Solutions, valantic Business Analytics
IBM Champion – IBM watsonx.ai, IBM Cloud, Enterprise AI