Any software generated with the very limited information accorded by a screenshot is hardly worth the time to comment on.
sounds like a new version of screen scraping to me.
Somebody will have to generate production quality screens and configure the workflow and nuances. Perhaps a little scripting involve to make it do just right.
The end result will still be a software. A screen-driven programming, just like a myriad of *-driven programming languages and frameworks we have now.