Why BGP, and when not
BGP exists because the internet isn't one network, and the question it answers isn't "what's the shortest path?" but "which networks will I exchange routes with, and on what terms?". That's why it's not the right tool everywhere, and why it's the only tool for some jobs.
Inside one organisation, an interior gateway protocol (IGP) such as OSPF is usually the better fit. Every router is yours and trusted, the goal is the fastest, shortest path, and the protocol is designed to converge quickly with little configuration.
Between organisations, the priorities change completely. You don't trust the other side's routes by default. You want to decide which paths to prefer for business reasons (cost, contracts, performance), not just hop counts. You need to carry a million routes stably. BGP is built for exactly that: it's a protocol of policy, where every session has explicit rules about what's accepted and what's announced.
When you need it, and when you don't
| Your situation | The simplest thing that works | Move on when… |
|---|---|---|
| One site, one internet connection | A default route to the provider | You add a second provider |
| A few sites, a provider-run WAN | Static routes, or the provider's service | Sites change often, or you need automatic failover |
| Several sites with backup links | An IGP (OSPF) inside | You connect to other organisations |
| Your own addresses, two or more providers | BGP with each provider, default routes or full tables in | You become a provider, or join an internet exchange |
| A provider, data centre or exchange member | BGP everywhere it meets the outside, an IGP inside | (You're there) |
The pattern: each step is taken because a specific failure or limitation forced it, never because the bigger design looks more professional. If you can't name what a piece of the design protects you from, it probably shouldn't be there.