I will address the question in the case $f(z) = (-1)^z = \exp(z \log(-1))$ and $A$ is a finite-dimensional commutative (associative unital) $\mathbb{R}$-algebra, where $\log(-1)$ is a suitable choice of the logarithm in $A$.
Wlog. $A$ is a local $\mathbb{R}$-algebra $(A,\mathfrak{m})$. There are two possibilities. First possibility is $A / \mathfrak{m} \cong \mathbb{C}$, in which case $A$ has a $\mathbb{C}$-structure extending the $\mathbb{R}$-structure, and the second possibility is $A / \mathfrak{m} \cong \mathbb{R}$. Either way, $A \cong \mathbb{K} \oplus \mathfrak{m}$ as vector spaces (not as algebras though!), where $\mathbb{K} \in \{\mathbb{R},\mathbb{C}\}$.
1. Claim: $\log(-1)$ exists iff $A$ has a complex structure.
Proof: Write $w = r + X \in A$ for $r\in \mathbb{K}$ and $X \in \mathfrak{m}$. We need $\exp(w) = -1$, i.e. $\exp(X) = -\exp(-r) \in \mathbb{K}^\times$. However
$$
\exp(X) = 1 + \sum_{k=1}^n \frac{X^k}{k!} \in 1 + \mathfrak{m} \subset A^\times,
$$
where $n \in \mathbb{N}$ is the smallest integer such that $X^n = 0$, as all elements of $\mathfrak{m}$ are nilpotent. Hence, in particular
$$
-\exp(-r) = 1,
$$
which is possible iff $r \in \mathbb{C}$. Moreover, we have $X = 0$ for the following reason:
$$
0 = \sum_{k=1}^n \frac{X^k}{k!} = X \left( 1 + \sum_{k=2}^n \frac{X^{k-1}}{k!} \right),
$$
but the expression in the brackets is a unit because $\sum_{k=2}^n \frac{X^{k-1}}{k!} \in \mathfrak{m}$. This argument is important for our second claim.
Thus the function $(-1)^z$ makes sense only if $A$ has a complex structure.
2. Claim: there is no $z \in A$ without complex components such that $(-1)^z \in \mathbb{R}$
Proof: Fix a value/branch $\lambda:= \log(-1) \in \mathbb{C}^\times$. Since $z$ has no complex components, we have $z = X \in \mathfrak{m}$, i.e. $z$ is nilpotent, and $(-1)^z = \exp(\lambda X) =: \exp(Y)$ for $Y:= \lambda X \in \mathfrak{m}$. Same argument as above shows that $\exp(Y) \in \mathbb{R}$ if and only if $Y = 0$.
Thus there aren't any non-trivial $z \in A$ such that $(-1)^z \in \mathbb{R}$.
I suspect one could show something similar in the case of infinite-dimensional commutative (associative unital) Banach algebras.
I don't have an answer for non-commutative algebras, finite-dimensional or not.