Computer networking is a hard problem. Partitioning the problem into layers simplifies it into easier problems. Also, because the layers have well-defined interfaces (ways to communicate) it is straightforward to change things.
Example: Ethernet uses a ‘physical layer’ to understand the voltage pulses in wires. You can swap to a different ‘physical layer’ if the signals are carried on a different kind of carrier - e.g. ethernet coax or twisted pairs. Changing to WiFi will need different physical and link layers but the rest of the stack of layers can stay unchanged.
Example: There is a spec for ‘IP over avian carriers’, a physical layer protocol using carrier pigeons. It was implemented and worked but slowly and unreliably. The spec has been updated to allow better reliability and further updated for IPv6 (though they needed changes to the link layer too.)
Example: The ‘physical’ layer doesn’t need to know whether the data stream is encrypted or not. That is handled by a different layer. So the development of the physical layer solution is simpified to just handle the signals and pass on the bit-stream to the next layer.
There is a ‘standard’ networking model using 7 layers. Common implementations often merge or omit some layers but the overall concept remains.