The math doesn't matter if the trust isn't there
In 20 years leading supply chain transformations, from SAP APO to o9 Solutions, I've seen one pattern repeat itself: adoption dies in the "black box".
We can implement the smartest AI models available, but if demand planners don't understand why the forecast changed, they override it, and the value is lost.
I've been evaluating the o9Predict AI architecture recently, and the concept of "Forecast Tournaments" stands out as a practical fix for this trust gap.
There is a key architectural nuance here: it's not just running models blindly every week. It follows a disciplined lifecycle.
- The tournament (strategy). Periodically, the system runs a competition to test multiple algorithms against history and select a "Champion" for every product group.
- Train and forecast (execution). In periodic cycles, the system retrains that specific Champion on the latest data and generates the forecast.
From black box to glass box
This moves us from a "black box" that just outputs a number to a "glass box". The system shows exactly which recipe won the tournament and why, for example: "Random Forest won because it caught the price drop pattern."
That transparency is the key to user adoption. When teams see the logic behind the Champion, they stop fighting the system and start managing exceptions.
It's shaping my perspective on future roadmaps, and on how we drive true touchless planning maturity.