i think that the cucumber/gherkin stuff she talks about is part of a set of tools for this kind of thing.

you describe behaviours, and then use that to drive tests and implementation