Let $\mathcal{A}$ be an invertible operator on a finite-dimensional vector space $V$ of dimension $n$. Let $g(t)= a_0 + a_1 t + a_2 t^2+ ... + a_n t^n $ be a characteristic polynomial of $\mathcal{A}^{-1}$. Then, from Cayley-Hamilton theorem, we obtain: $g(\mathcal{A}^{-1})= a_0 \mathcal{I} + a_1 \mathcal{A}^{-1} + a_2 (\mathcal{A}^{-1})^2 +... + a_n (\mathcal{A}^{-1})^n= \mathcal{O} $, where $\mathcal{I} $ and $\mathcal{O}$ are an identity and zero operator, respectively. Hence, $ (\mathcal{A}^{-1})^n = -\frac{ a_0}{a_n} \mathcal{I} - \frac{ a_1}{a_n} \mathcal{A}^{-1} - ... - \frac{ a_{n-1}}{a_n} (\mathcal{A}^{-1})^{n-1} $. If we compose both sides of last equality with operator $\mathcal{A}^{n-1}$, then we get:
$ \mathcal{A}^{-1} =-\frac{ a_0}{a_n} \mathcal{A}^{n-1} -\frac{ a_1}{a_n} \mathcal{A}^{n-2}- ... - \frac{ a_{n-1}}{a_n} \mathcal{I} $.
Is this reasoning correct, can we express $\mathcal{A}^{-1}$ in terms of $\mathcal{A}$ this way? Thanks a lot in advance.
Yes, your reasoning is absolutely correct.
You could obtain the same result by using the characteristic polynomial $p_A(t) = \det ( A-tI)$ of $A.$ It is even a classical method, when the constant coefficient of this $p_A$ is nonzero, to prove that $A$ is invertible and compute its inverse.
Moreover (just for fun), note that $$p_{A^{-1}}(t) = \det (A ^{-1}-tI)= \det(-tA^{-1}(A-t^{-1}I))=(-t)^n\det(A^{-1})p_A(t^{-1}).$$