How common is it for researchers to use computer algebra systems? I'm working in electrical engineering and moving into quite theoretical areas. For example, I'm reading right now some papers on tail bounds / concentration of measure in probability. They pull almost out of nowhere some very clever functions that make the inequalities work in a nice way.
For example, they write
$\frac{b}{b - a}e^{ta} - \frac{a}{b - a}e^{tb} = e^{g(u)}$
Where $g(u) = -\gamma u + ln(1 - \gamma + \gamma e^u)$, $u = t(b - a)$, $\gamma = \frac{-a}{b - a}$
This all works out nicely, but it would be certainly much harder to go backwards. This is also a very simple example compared to others I've seen.
I can understand coming up with this through a trial and error process. Or perhaps they are using standard tricks I'm just not aware of yet. But, how can one know that such a function $g(u)$ even exists, especially in more general situations?
Is it possible to use a CAS and ask it "write this function in the form $e^{g(u)}$, or tell me it's not possible"? Is a CAS useful in general? Is it common to make use of them?