Agile, ML and Agents strategies for revenue generation

How to introduce ML in your company?

That is the question that I was asked last Friday. I found myself saying "it is tricky, ML typically comes in a right-angle to the people process, and that does not work!" Right-angle? You might say? And you would be right! Therefore I picked up a deck I wrote last November, kept only the key statements, added the context, and recorded a 17m presentation here: https://www.youtube.com/watch?v=08uEBKft_Bg  which will not tell you "why" but "how", as in "how to introduce machine learning in your company".

The finer story here, and not highlighted in the presentation is success with or without machine learning is first about your team maintaining maximum performance throughout the project. See it as a rocket that is maintaining maximum thrust. And for that to happen you must carefully ensure that everyone can invest their maximum with a full understanding of what they and the team wants to achieve. Teams may sometime need help to grow this culture and vision of how good they can be.

Below I highlight some of the key slides and statements from the presentation.

1) Use agile processes

Agility is about deconstructing you processes. ML strives on "simple". Furthermore ML strives on simple and fast iterative processes. These can only be provided with an agile approach.

2) An agile team can incrementally bring in machine learning 

You can be running a process such as Scrum, and incrementally grow your ML usage.


3) Agentic ML has its limits

As of April 2026, people are still better off owning requirements and quality in software developments, ML agents are simply not sufficiently capable of "reading one desires" and really be thorough in ensure proper quality in software.

4) Alignment in process does not ensure success

The diagram below illustrates how both agile and ML processes can be aligned, and yet the people and the machine learning output can be totally unaligned! 

5) Staggered process step ownership is one way to align people and ML

Each tier of the agile process process is either owned by people or machine learning, except tasks where both people and tasks have independent tasks.

6) Full ML agent with ownership of analysis, tasks and drive is another approach

People maintain ownership over leadership and validation. The scaling of machine learning and flattening the requirement and design needs help ensure the success of the project.


All original content copyright James Litsios, 2026.