A productivized scraping service - useful! Entire companies are built around scraping certain popular sites - this is a disruptive idea indeed. A growing catalog of up-to-date scrapers for popular websites would put of lot of freelancers out of work. I would invest in this.
However, the ML claim is highly suspect. There is no way that a machine could reliably understand the semantic content of a website - that would require Artificial General Intelligence. If anyone could do that, it would've been Google. But even Google relies on human-edited structured metadata to define the content of sites (i.e. Rich Snippets and the like).