Is it problems with ETL, lack of labelled data, shortage of data scientists, disconnect between research, engineering and management or something else entirely
Customers don't understand what you're talking about and they don't like blackbox solutions.
I.e. ML-heavy offerings are throwing ML-rich buzzwords while customers want to solve problems who "talk" their language.
If solutions were demonstrated, do you think customers would still care about black box vs understandable?