Well, 0th order approximations are fine if the gradient is indeed about zero for all intents and purposes. It's just very much not the case for exponential functions.

Maybe your analogy could work if you want to say that exponential growth might feel like linear growth locally?