$(\Rightarrow)$ Let $(x_n)$ be Cauchy w.r.t. $\|\cdot\|_A$. Then $(x_n)$ is Cauchy w.r.t. $\|\cdot\|_X$ because $\|\cdot\|_X\leq \|\cdot\|_A$. Thus, there is $x\in X$ such that $\|x_n-x\|_X\to 0$ (because $X$ is Banach). As $A$ is bounded, it follows that $\|Ax_n-Ax\|_Y\to 0$. This implies that $x_n\to x$ w.r.t. $\|\cdot\|_A$. So, $\|\cdot\|_A$ is complete.
Remark. The fact that $x_n\to x$ w.r.t. $\|\cdot\|_A$ follows from your estimate with $x_m$ replaced by $x$. This is the main point: the boundedness of $A$ should be used to prove that convergence in $\|\cdot\|_X$ implies convergence in $\|\cdot\|_A$. So, we have to deal with the diference $x_n-x$ instead of $x_m-x_n$ (as in your trying). The fact that Cauchy in $\|\cdot\|_A$ implies Cauchy $\|\cdot \|_X$ follows from the definition of the norm and thus it has nothing to do with the boundeness.
Remark 2: The converse follows from the Open Mapping Theorem (see here):
$$\|Ax\|_Y\leq \|x\|_A\leq c\|x\|_X.$$