By Kaushik Rajan
Publication Date: 2026-03-07 13:00:00
Forecasting approximately $50 billion in advertising revenue using econometrics, time series modeling, and causal inference. When a senior vice president asked how certain we should be about a number, I couldn’t give him a point estimate and shrugged. I had to quantify the uncertainty, trace the causal chain, and explain which assumptions would destroy the forecast if proven wrong.
None of these works involved a Large Language Model (LLM). None of this could have happened.
If you’re a data scientist feeling let down by the AI wave, this article is a reorientation. The skills the industry is giving up are the very ones that are becoming scarcer, more in demand and better compensated. While everyone else chases the next base model, the market is quietly reassessing fundamentals.
This article describes five specific skills (I call them the anti-hype stack), explains why each of these skills resists automation, and gives you a 90-day roadmap for building these skills. But first, a quick look at why the hype…
