The digital assistants don't have recorded phrases, it's speech synthesis. Someone once speaks a lot of text that contains all the different sounds that make up words, and that's then sliced and the input for the synthesis. A lot more work than just recording a small set of phrases.
And of course a persona is more than the pure voice, but also writing and how they use their voice to say things.
EDIT: an article about Siri's voice actor: https://www.vox.com/2015/6/23/8831131/siri-voiceover-susan-b...
> "I recorded four hours a day, five days a week for the month of July," […]
> The goal is to try to include every possible sound (usually drawn from a syllable-long building block) so they can be assembled in every possible combination for every possible word. To record these, voice actors are forced to recite gibberish-like sentences that include all of the English language's different sounds […] including absurd phrases like "oil your mills jewel weed today."