Originally published on AI Tech Connect.
What changed, what did not, and who should care Two teams have the same problem in different shapes. A logistics operator in Pune needs next-week volume forecasts for 40,000 lane-and-SKU combinations so that vehicles get positioned before demand arrives rather than after it. A utility in the north of England needs half-hourly load forecasts for a few hundred substations, where being wrong by a small margin at the wrong moment costs real money on the balancing market. Both teams have heard that time-series forecasting now has foundation models, that you can point one at a series and get a forecast without fitting anything, and both are wondering whether that makes their existing pipeline obsolete. For the Pune team, the answer is close to yes and it is a genuinely large win. For the…
Top comments (0)