Yes you can create your tables with ON UPDATE CASCADE foreign keys, but are those really the only places the ID is used?
Sometimes your own data does present a natural key though, so if it’s fully within your control then it’s a decent idea to use.
For example, suppose you have an information management system where the user can define their own logical fields or attributes. Naturally those names should uniquely identify a field. That makes them an easy candidate for a natural key. But I would still use a surrogate key.
I've worked in two systems that had this feature. In one the field name was a primary key, and in the other they used a surrogate key and a separate uniqueness constraint. In both a requirement to let users rename fields was added later. In the one that used a surrogate key, this was an easy and straightforward change.
In the one that used field name as a natural key, the team decided that implementing the feature wasn't feasible. The name change operation, with its cascade of values needing to be updated for all the foreign key relationships, would have been way too big a transaction to be doing on the fly, in a live database, during high usage times of day. So instead they added a separate "display name" field and updated all the queries to coalesce that with the "official" name. Which is just ugly, and also horribly confusing. Especially in oddball cases where users did something like swapping the names of two fields.
On the other hand, I remember a CEO wanting to be #1 in the database, so even non-natural primary keys can change... haha.
I'm curious, where else would they be used?