AI, ML and Deep learning require  data to be prepared a very specific way before it can be used to answer business questions.  The data needs change depending on the  question being asked or the data model/library being used.

1. AI, ML and deep leaning is only as good as the underlying data. Get the data right and the answers are better.

2.  Why pay data science salaries to do data engineering work like cleaning up the data, joining the data, adding new data sources etc.?

3. Retention:  A good data scientist is a difficult find and hire and they leave if all they do is data engineering work.