Our view is that it is difficult to perform wholesale replacement of a data scientist with an AML framework, because most machine learning problems require domain knowledge and human judgement to set up correctly.
Automated Machine Learning — A Paradigm Shift That Accelerates Data Scientist Productivity @ Airbnb
Hamel Husain

I strongly agree with it. In fact, data scientists need new tools to help their work and AML approaches might be a valuable choice when building a model. As you said, sometimes the domain knowledge is required. For instance, to discover an important non-trivial feature of the model, the domain specialization is welcome. Repeating a cliche, there is no silver bullet.