For everyone interested i solved the problem by rethinking the whole algorithm. Now when the first set of actions starts the level blueprint spawns an actor with the random delay and an event tick in the level blueprint checks if a boolean variable in the countdown actor is true. At the end of the set of delays, when the character stops talking a branch checks that variable, if it’s true it means the countdown has stopped and it goes back to the random sentence part, if it’s false it goes on with the user’s answer.
Thanks everybody, hope this helps, sometimes you just need to stop and think