You need HKTs to have type-checked generic monads. Almost no languages have them.
Dynamic languages can implement generic, but not type-checked monads. Many static languages can implement type-checked but non-generic monads.
Now you need to show how type-checked generic monads are critical. That's much harder than showing that the monadic pattern is useful.
In practice, 75% of the time that I use monads I am using some kind of monad stack. I'm sure you can argue that this is unnecessary or that I only do so because Haskell affords it... but on the other hand, I believe that transformer stacks do a good job of elaborating the space of effect types with an efficient, expressive language. Without them you either elide side effect reification or build a lot of one-off monad stacks.