Apple's AI email categorization feels outdated already

Apple recently enabled AI categorization of emails into four categories: Primary, Transactions, Updates, and Promotions. Usually, when a company adds an AI feature, I scramble to find the off-switch, but I’d heard so many negative remarks about it, I just had to see for myself. My expectation to find myself disappointed by Apple’s design choices and technology was met.

About the categories

For starters, let’s just pretend like the feature works well and does what it’s supposed to do. There is the choice of categories—Why? If we are using AI, why should Apple make the choice what categories my mail gets sorted into? Isn’t the point of using AI that it is flexible, and, you know, intelligent enough to come up with categories that are useful for the kind of mail in my inbox?

I see two ways this could have been done better by Apple. One is that true AI categorization of emails should go through all of them and then come up with useful categories for the emails it finds. Another is that it should allow me, the user, to select a group of emails, which, I think, should be in one category, and use that input as a starting point for a new, user-defined category. Instead, Apple decided that I would need four categories they selected.

I think this shows a fundamental problem: A lack of understanding of AI by the product designers at Apple, and a lack of empathy with users. I’ll elaborate:

Lack of understanding of AI

What I mean by that is the way of thinking this choice of categories represents. It’s an old-fashioned solution; The idea to sort emails into categories has been around for years—Gmail started doing it in 2013. As far as I know, the original problem Google aimed to solve was a flood of newsletters drowning out important emails. The solution was to push newsletters into a separate tab—awesome, that’s actually helpful.

Over time, Google added more categories. And now, Apple has come up with its set of categories to sort emails into, and solved the sorting process with AI. Apple’s designers didn’t consider that AI is capable of coming up with individual categories for each user. Instead, they defined categories, then left it to the engineers to come up with some Machine Learning algorithm to sort emails into those categories.

The feature feels outdated while it’s still in beta because the product designers didn’t utilize AI’s capabilities properly. It feels like building a tunnel in Vegas for self-driving cars to solve traffic, instead of a maglev train, a subway, a tram, or even a cable car… but I’m getting side-tracked.

A lack of empathy with users

The entire AI-hype lacks empathy for users, and so does Apple’s Apple Intelligence push. It’s the job of product designers to say no to features and products more often than yes—hell, a thousand noes for every yes was Apple’s WWDC intro movie in 2013! (Vimeo.com)

I’m probably biased here, but almost every single person I’ve talked to in the past six months either does not care about AI, or dislikes it. Even my technologically challenged aunt asked me how to get rid of the Meta AI in her WhatsApp!

The only people I’ve talked to who like to use AI can be sorted into two categories: People curious about tech, who understand the risks, and want to experiment with it, and those who forward SPAM emails to their nephews to let them know that a Nigerian prince is about to send them a million dollars.

Pushing AI on your users before it’s ready for prime time and while they don’t care about it or dislike it, demonstrates a lack of empathy with your users. Forcing designers to push these technologies when they know it’s a bad idea is bad management.