The pitch runs forty minutes. The model handles the sample data cleanly, answers the planted questions, and someone says the word “transformative” before the deck finishes. Three weeks later there’s a pilot, a budget line, and a delivery date. Nobody has written down what the pilot is meant to prove.
Not out of carelessness. The demo already looked like proof. Saying otherwise would have meant admitting the excitement arrived before the analysis did.
The architecture review does happen, eventually. Not before the decision. After it — once there’s already a budget line and a delivery date, and the only question left is how to make it work, not whether it should. The review that could have caught the gap wasn’t absent by accident. It was invited too late to matter, on purpose.
The demo answered one question: can it do this. Nobody asked the harder one — is doing this worth what it costs — because the first answer already had everyone convinced. The pilot was never mapped against the value chain it was supposed to sit inside. The non-functional requirements that would have surfaced the volume, the latency, the edge cases the demo never touched were never written.
Six months later the pilot is technically successful and worth nothing anyone can point to. The postmortem calls it an AI problem. Wrong model, wrong vendor, not ready for scale. It was never an AI problem. A demo is not a business case. It only looked like one because nobody in the room could afford, politically, to say otherwise.
The model did exactly what it was shown doing. Nobody checked if that was ever the actual job.
— Signal / unfiltered observations from the field. Fortnightly.
#OrdoAnimi #EnterpriseArchitecture #ArchitectureLeadership #AIStrategy #DigitalTransformation #ITStrategy #Execution #CIO