Which fields are most and least likely to be impacted by AI?

Certainly AI will affect how much coding we do by hand. The actual data science part is harder to automate, because every problem requires business context and an understanding of how to achieve your goal with the data you have.

That being said, as someone who has concentrated heavily in one niche (forecasting), I am curious which fields in DS/ML people think are most or least likely to be automated substantially by AI. Forecasting, Optimization, A/B testing, Causal Inference, Vision, Anomaly Detection, etc?