API gateway patterns: edge vs service mesh — when to use which?

We're currently running Kong as an API gateway at the edge of our infrastructure, primarily for rate limiting, authentication, and routing external traffic. Now, our internal teams are looking at adopting Istio for inter-service communication within our Kubernetes clusters, aiming for features like mTLS, traffic management, and circuit breaking. I'm seeing a lot of feature overlap between Kong and Istio, particularly around load balancing and circuit breaking. How do other companies divide responsibilities between an edge API gateway and an internal service mesh? Is it common to have both? What's the recommended layering or delineation of concerns to avoid duplication and minimize operational complexity? I want to ensure we get the benefits of both without creating an overly complex or redundant network architecture.