Solve enterprise search. Reposition consumer offering around strengths in increasing open data world derived from enterprise. Rethink search UX to vertically integrate bookmarking, analytics and note-taking, going beyond discovery. Make search results input to further processing in Wolfram-lite ways. Make the UX unsocial (Google's social flavoring to promote Google+ is a weakness... social discovery is happening through 'like', RT and '+1' type actions which amplify status-signaling and fun content over useful content, so social search isn't really working). Use enterprise stronghold to invent a model of 'expert' search. When I search for (say) a recipe, I'd like an expert chef's behaviors to modulate the results more than amateur's.

Google is definitely vulnerable in the next 5 years. Question is only whether Bing can go beyond fast follower tactics and leapfrog.

Search doesn't _look_ broken because we've been conditioned to move blame elsewhere. It has its current silo form due to considerations unrelated to the natural logic of discovery (advertising, social). Google has gotten complacent about pushing the envelope (otherwise they'd have solved the perennially broken bookmarking problem by now).