Consider the One-Time Pad (OTP). Suppose two parties, A and B, generate a completely random secret key in a real-life meeting, and they keep this secret. Now A wants to send some message to B using the OTP. The point is that in the OTP we can use any key only once. So we have to come up with a system to share a newly chosen key in a secret way.
Why cannot A just concatenate a new randomly generated secret key (needed for the next communication between the two parties using the OTP) to this message?
For instance, A could concatenate this new key at the beginning of the message. (Supposing they have agreed upon a fixed key-length beforehand, so that B is able to tell which is the new key and what was the "actual" message.) In this way A and B would agree on a new secret key which they can then use for their next communication.
Probably this is insecure, but I do not immediately see why…