when to use event sourcing vs. a regular database?

hey folks, i've been thinking a lot lately about how we store and manage state in our systems. right now, we're mostly using a pretty standard relational database approach for most things. but i'm curious about when event sourcing really makes sense as an alternative or a complement. besides just audit trails, what are the key architectural benefits you've seen, and what are the real-world trade-offs compared to traditional databases? i'm especially interested in opinions on complexity and operational overhead.