I've used a lot of both. Excel is much better for most tasks where you have < 1mil rows of data. Its easier to look at the data, easier for novices and fast enough. Just being able to scroll through the data is very valuable just to get a feel for it. The biggest drawback is VBA, if you could write excel macros in Python it would be a hit.
If you have more data, Notebooks can handle that better. However I've noticed lots of colleagues skipping notebooks and using IDEs instead. Much easier to work with and better for scm. I'm not a huge fan of notebooks any more.
Now that's quite a generalization... most tasks? If all you're doing is a=b+c then perhaps. I work in HFT and even for trivial data exploration I would never even consider touching Excel; why would I? Even if it's just 100 rows. No thanks. Once you're comfortable with Python / its scientific stack, the exploratory part of data analysis becomes fast and trivial.
What I would like to see is notebooks becoming more IDE-like. This is already happening gradually, e.g. with JupyterLab replacing Jupyter notebooks ([1])
If you have more data, Notebooks can handle that better. However I've noticed lots of colleagues skipping notebooks and using IDEs instead. Much easier to work with and better for scm. I'm not a huge fan of notebooks any more.