"Load balancing" decisions

Chatting with some colleagues yesterday, one of them used the phrase "load balancing" when talking about how you could combine computer-based and human-centric decisioning. The context was a discussion around Deep Blue, IBM's famous chess computer, and how much processing power it needed to win chess matches against the top human player - 256 processors and 200,000,000 positions evaluated a second. This led on to another discussion, one somewhat inspired by my recent reading of Blink and by Larry Rosenberger's presentationon a similar topic, whatif we tried to build a computer to assist a player not replace them completely? In EDM terms, what if we:

• Used analytics to score potential moves for risk and upside
• Used automation to eliminate many of the possible moves that are clearly unsuitable
• Automated those moves where there is an obviously best move
• Left the less clear ones for a player to make but gave the player lots of extra information about why they were being asked

How much processing power would this take? Is this kind of "load balancing" interesting or useful? It was a fascinating discussion, enriched by the wonderful concept of "load balancing" between people and computers.