Three Phases I Did Not Plan: What Two Years of Internal AI Posts Revealed

A Pattern I Did Not Notice Until Someone Else Pointed It Out

For roughly two years I posted internally, on and off, whenever I learned something about AI worth sharing. Nineteen of those posts turned out to be about AI or SAP BTP specifically. I never once sat down and planned a curriculum. Only when someone reviewed the whole archive at once did a clear three-phase progression become visible, one I had walked through without noticing I was walking through it.

Why Post Internally at All

The habit started from a simple belief: explaining a topic is the fastest way to find out what you do not actually understand yet. Viva Engage, an internal Microsoft channel available at Accenture, was the natural place to do that, in front of an audience of SAP practitioners, AI learners, and colleagues exploring the same generative AI wave I was.

The Three Phases, in Hindsight

Phase one was fundamentals. Early posts explained core AI concepts in plain language: machine learning as a statistical revolution, neural networks as a shift from explicit rules to learned patterns, natural language processing as teaching computers to work with human language. These were written to be accessible, not technical showcases.

Phase two was hands-on experimentation. The focus shifted toward practical learning: working inside SAP AI Core and SAP AI Launchpad, provisioning free-tier AI services on SAP BTP, running Jupyter notebooks against HANA data, building one simple neural network from scratch just to see it work. These posts documented discovery as it happened, including the parts that did not work on the first try.

Phase three was enterprise architecture. Later posts moved toward multi-agent AI architectures, SAP AI Foundation, MLOps patterns for generative AI, and SAP Joule evaluated from a technical consultants