You actually made a lot of arguments in favor of an ORM already.

It's not about not having or wanting to write sql, even though this will still allow people to use a database that do not know sql at all. Let's not focus on that however.

Developing becomes a lot more straightforward and faster because you can focus on your objects instead of queries and parsing. Most people are not doing anything complicated (sql wise) and a good orm will make the same queries as you would. This removes a lot of boilerplate sql code.

You also would not want to litter your code with sql statements that make it harder to read. So you can choose to abstract them into their own library or something, but now you still have half an orm that you had to build and expand for every database interaction that you intend to do. It's basically the same reason why you would use templates. Nobody wants to do their formatted output in loads of print statements.

Because you do not have to use sql all over the place remodelling parts of your datamodel becomes a lot easier and more consistent.

I am definitely biased because i have been using ORMs for at least the past 10 years. In all those years at various clients using various ORMs i only once had to revert to writing a raw sql statement.

Most of my time i've used Django's ORM (python) and SQLAlchemy (also python) but also used Ruby on Rails' Active Record. If queries were slow they all offer tools to debug what is going on. Every time except once i have been able to change the ORM statements a bit to optimize.

The only time i would not use an ORM is either if the programming language or framework used would not offer a decent/mature ORM or if performance was the only important metric and could not be solved otherwise. Maybe in that case you would also not want to use python or ruby to begin with. In all other cases spend developer tjme is often much more expensive then spend querytime.