Do we really need predictability in Agile and Digital world?

AI and machine learning can only predict or decide based on the provided data. If you give them little or no data, don't expect AI and machine learning to fill the gap. (The opposite is also a problem: too much data can exacerbate known biases.)

Agile projects usually mean we start with very little data, therefore we do the minimum amount of prediction and decisions.

Prediction in an agile project is like a sailboat that constantly corrects course as it unfurls more sails and tacks into the wind.

1) Agile is predictable in that it will provide value frequently, rather than waiting to the end of the project.

2) Agile shouldn't be used as an excuse for poor planning. You can still have a solid business case, high level requirements, a fixed budget, requirements definition, well-defined scope of work, etc., and still implement the project using an Agile approach. It doesn't mean you start without a good plan, it just means that you value change over following the plan. You embrace change. You don't start off without a clear idea of where you're headed. You may change your mind about how to get there, or even adjust the final destination somewhat, but you don't just head off not knowing where you're going.

And the whole notion of story points and velocity is about being able to not only improve performance, but forecast cost and schedule to complete the project.