The usefulness is a function of how quickly the consequences from poor coding arrive and how meaningful they are to the organization.

Like in finance if your AI trading bot makes a drastic mistake it's immediately realized and can be hugely consequential, so AI is less useful. Retail is somewhat in the middle, but for something like marketing or where the largest function is something with data or managerial the negatives aren't as quickly realized so there can be a lot of hype around AI and what it may be able to do.

Another poster commented how very useful AI was to the insurance industry, which makes total sense, because even then if something is terribly wrong it has only a minor chance of ever being an issue and it's very unlikely that it would have a consequence soon.