Here's one approach. (It is essentially the same approach as @Max's, presented in a different way.) It is based on the notion of quotient of a language $L$ by a word $x$:
$$ x^{-1} L = \{ w \mid xw \in L\} .$$
If $L$ is regular, so is $x^{-1}L$, because it is the language accepted by a DFA $D$ for $L$ modified so that the initial state is the state reached by $D$ after reading $x$.
If $\epsilon \not\in L$, your $L'$ is simply
$$ \bigcup_{a \in \Sigma} a^{-1}L \cdot a \enspace. ~~~~~~~~~~~~~~~~~~~~~~~~~(1)$$
Hence, it is regular because regular languages are closed under concatenation and finite unions. If $\epsilon \in L$, then the definition of $L'$ should be clarified. Suppose it means that $\epsilon$ is also in $L'$. (There is no first letter, hence there's nothing to do to $\epsilon$.) Then it's enough to add $\{\epsilon\}$ to $(1)$. If $\epsilon$ is excluded from $L'$, then $(1)$ is the result.