An alternative proof: If the claim is false then
$$ \tag{*}
e^{\lambda z} - p(z) = q(z) e^{h(z)}
$$
with non-zero polynomials $p, q$, and an entire function $h$.
To simplify the notation let us assume that $\lambda = 1$, this is no
loss of generality. Then
$$
e^{h(z)} = \frac{e^z - p(z)}{q(z)}
$$
and for $|z| = r$ sufficiently large, some $m \in \Bbb N$, and positive real constants
$C_1, \ldots, C_4$
$$ e^{\operatorname{Re}h(z)} =
\lvert e^{h(z)} \rvert \le C_1 (\lvert e^z \rvert + C_2 r^m)
\le C_1 ( e^r + C_2 r^m) \le C_3 e^r
$$
and therefore
$$
\operatorname{Re }h(z) \le C_4 + r
$$
for $|z| = r > R_0$. So
$$
A(r) :=\max_{|z|=r}\operatorname{Re}h(z) \le C_4 + r
$$
Now use
to estimate the Taylor coefficients of $h$
in terms of $A(r)$, and conclude that $h$ a polynomial
of degree at most one: $h(z) = az + b$.
Finally, substitute $h$ back into equation $(*)$ and compare the
growth of the different terms.