Everyone agrees that cryptography should be replaceable. Few organisations can say how long a replacement would take, which is the only version of the question that matters.
Start with an inventory
You cannot replace what you cannot list. The useful inventory is not of algorithms but of places where an algorithm is pinned: firmware that ships with a fixed cipher suite, an appliance whose key exchange is not configurable, a partner integration where the algorithm is written into a contract.
The awkward entries
The difficult items are rarely the ones in the data centre. They are the field devices with ten year replacement cycles, the leased line where the encryption is the carrier responsibility, and the archive whose contents must remain confidential well past any current algorithm lifetime.
Separate key supply from key use
The single change that buys the most agility is putting a key management layer between whatever produces key material and whatever consumes it. Once encryptors take keys over a standard interface, the source can change without touching the encryptors. That is what makes it possible to add a quantum key source later without a forklift upgrade.
Decide what is worth protecting for how long
Not all traffic needs the same horizon. Session data that is worthless in a week does not justify the same investment as records that must stay confidential for thirty years. Sorting traffic by required confidentiality lifetime turns an unbounded programme into a short list of routes.
Those routes are where quantum-safe infrastructure earns its cost, and they are usually a much smaller part of the network than the first conversation assumes.