If you've recently called a company on the phone, you have more than likely experienced the automated machine that relays "...please press 1". It's even worse when it asks you to explain your problem, and then gets it terribly inaccurate, which is why banks are still full of people trying to get their account re-instated.
To understand why this is the case, let's break the reasons down into their respective categories:

Financial - The Return-On-Investment (ROI) just isn't there (yet)
The biggest reasons A.I was implemented in the first place was to save costs from paying employees (along with whatever 'run-around' answer the CEO would give). Ironically, that's exactly what the A.I companies hoped for - they provide their technology for 'cheap' so that they could train the models and make companies reliant on them, and then jack up the price when the timing was right. So the pricing model has done a complete 180 - it's now more expensive to have A.I implementation with specialists to train them, than to get the employees back to their desks.

Investors of these companies were cashing in on this growth, but because the ROI will not only take many, many years for it to make sense, an IBM CEO study also found that only around 25% of AI initiatives deliver expected ROI. Yikes, not something investors are willing to gamble on currently.

Technological - A.I is friendly, but not a friend
In regards to replacing the understanding and empathy of a human, A.I is not close. Whilst it may be agreeing, useful and knowledgeable, A.I lacks the genuine nature of human connection and understanding. This is why Teki exists - to build that bridge between tech knowledge and interpersonal empathy. 

Legal - What's ethical or not is still... questionable
If the companies themselves are not using A.I, their employees most certainly are. But the use of these external tools used without your permission seems questionable - if it helps solve your problem, but gives away your personal details, is this ok? 
Privacy policies of many companies need a revamp, and depending on how Government intervention occurs, who knows whether A.I usage will continue to be a 'free-for-all'.