I believe it is all about training the sales associates on what level of interaction is appropriate and how to read the customer. Even at higher-end stores, I see wide variation in the astuteness of sales associates. Knowing immediately where to go across an entire department to show me items that might suit my need is very helpful. It's not helpful for the associate to take over for me when I already am going through a stack of jeans looking for my size to do that same task for me while I just stand there watching. Unfortunately, even in high-end stores, my experiences skew more to the latter example.
Certainly dropping the local names had an impact. I recall the front page headline of Chicago newspaper "Nightmare on State Street" when the name change was announced and that sentiment was widely shared. The much bigger problem though, is the degradation of products and service. May gave up the attributes that distinguished Marshall Field's and the comparable stores elsewhere, replacing them with offerings closer to other mid-grade department stores.
If AI can successfully identify the attributes most important, it could be helpful but given the variability of attribute importance across different people, it would be difficult to achieve. A good option is to allow consumers to easily identify a group of acceptable substitutes (e.g., any other zero calorie cola). Currently, if this option is even available, it is cumbersome to use.
The commission-based pay structure needs a revamp. People shop differently today. If I research an item online, decide to purchase, check online for store availability, then drive to the store and purchase, the sales associate had nothing to do with the purchase except taking a minute or two to ring up the sale. But they are compensated the same as an associate who took a fair amount of time in-store to help me with my search. It's time to find a new way to compensate sales associates in many types of merchants.
Five-star reviews have become much less reliable than in the past. One issue is that five has become the expected rating for everything being up to expected levels. I've received countless solicitations from businesses requesting that I leave them a five-star rating if I am satisfied. A satisfied customer due to everything being at expected levels is a three, not a five. But if there is an average rating of three, most consumers would pass. The second issue is the number of planted reviews. Often on Yelp or Amazon, I see five stars and in the commentary notice a large number of reviews written just within days or even hours of each other, each giving five stars and the commentary basically says it is the best product, restaurant, etc. to ever exist. Obviously not real.
Some companies are doing it well, but not all. I see many investing heavily in the IT side, but not the people side. AI can go a long way, but it is a point of entry, not the final result. Too many are looking for an automated one-size-fits-all quick solution, but to be truly successful, analysis is an iterative process. The too often ignored need is the human contribution to make sure the right questions are being asked to create the necessary ROI, to understand that there probably isn't one silver bullet target, and to know how to further mine the data to get to the optimal result.