Noreply at Macy’s dot com asks,

How are you finding the product?

Good question. The last email from this company was to share this information:

Tracking: delayed

As far as I can tell from the site, not even the shipping company can find the product. Last scanned 10/22/2026, in Illinois.

This is a basic-ass customer-facing IT issue. One system is supposed to prompt customers to leave a review, or a one to five stars rating, or whatever, and another system is supposed to let people know when to expect their stuff to arrive. Right now, the two systems are not hooked up correctly, so I got asked for a review of a product that I haven’t received yet. no, this isn’t a problem of ha ha privacy nerd, you made your shopping experience worse by blocking tracking—I made an account and put in all my right info.

So what’s going on?

By activating data-driven and personalized omnichannel experiences, we empowered Macy’s to transform how customers experience their brand, creating seamless, magical shopping moments and increased relevancy, that’s what’s going on. Macy’s hired a fancy-pants consulting firm to set up “comprehensive customer journey experiences that feel personal, effortless, and magical.”

Which means, of course, less money left over for the basics.

Investments in Personalization—most of which is adversarial, getting the maximum price from those with fewer options and only giving discounts to those classified as bargain hunters—tend to crowd out investments in product and service improvements.

The optimistic view on AI shopping agents is that they will shift the relative ROI of the two kinds of investments. A retailer that chooses to splash out on a big “personalized multichannel experiences” project will be slammed by agents trying a variety of “customer journeys” to see which shopping persona gets the best deal. A retailer that fixes the basics will tend to get picked more over time, as more shoppers act like highly-informed shoppers thanks to agent help.

But another possibility is that shopping agents go the way of the Honey browser extension and start off working for their users, then sell out to retailers—who end up paying agents to choose them instead of improving.

I’m still cautiously optimistic because of what happened with Honey. A client-side tool doesn’t have the network effects that Google or Meta have relied on, to slide deeper and deeper into crime without losing users. People quit Honey without losing touch with friends. A shopping agent might even have reverse network effects, as a less widely used agent would face less investment in countermeasures.

More later.