The subgroup of $\operatorname{GL}_3\left(\mathbb{R}\right)$ generated by the matrices $$\left( \begin{array}{ccc} 0 & 1 & 0 \\ 1 & 0 & 0 \\ -1 & -1 & -1 \\ \end{array} \right)\,,\hspace{10pt}\left( \begin{array}{ccc} 1 & 0 & 0 \\ 0 & 0 & 1 \\ -1 & -1 & -1 \\ \end{array} \right)\,, \hspace{5pt}\text{and}\hspace{5pt}\left( \begin{array}{ccc} 0 & 0 & 1 \\ -1 & -1 & -1 \\ 1 & 0 & 0 \\ \end{array} \right)$$ is isomorphic to $A_4$.
When acting on $\mathbb{R}^3$, this representation does not preserve the unit $3$-ball (i.e. map the unit $3$-ball into itself). Given a $\mathbf{v}=(x,y,z)$ such that $\left\lvert\left\lvert \mathbf{v} \right \rvert\right \rvert<1$, we either have $\left\lvert\left\lvert A\mathbf{v} \right \rvert\right \rvert=\left\lvert\left\lvert \mathbf{v} \right \rvert\right \rvert$ or, up to permutation of the coordinates, $$\left\lvert\left\lvert A\mathbf{v} \right \rvert\right \rvert=\sqrt{x^2+y^2+(x+y+z)^2}$$ which achieves a maximum of $\sqrt{2+\sqrt{3}}$ on the unit $3$-ball.
Is there another representation of $A_4$ in $\operatorname{GL}_3\left(\mathbb{R}\right)$ that does preserve the unit $3$-ball?
Lord Shark the Unknown gave a fine answer. I want to supplement it with the following generalization.
Assume that you have a finite group $G$ acting on $V=\Bbb{R}^n$ by linear transformations. Let $(\ ,\ )$ be the usual inner product on $\Bbb{R}^n$. Let us define another bilinear form $\langle\ ,\ \rangle$ on $V$ by taking the average: $$ \langle x,y\rangle=\frac1{|G|}\sum_{g\in G}(gx,gy). $$ The following are easy to see:
Therefore the new metric given to $V$ by this averaged is invariant under the action of the group $G$. Call the resulting inner product space $V'$. All this immediately implies that balls in $V'$ are stable under $G$.
But, the space $V'$ is really the space $V$, we only changed the metric. You can easily find an orthonormal basis of $V'$. Rewriting the matrices of $G$ w.r.t. such a basis results in orthogonal matrices, i.e. $G$ acts by distance and angle preserving transformations.
Closing remarks: