AO is still new to the PA framework. Despite the random yield twist, his model essentially is a private cost case. That would be hard to sell. Change it to private risk, for which the private information is incontractible. Also, after gathering the information, the contract should also motivate effort investment. This is a moral hazard problem. So we need the limited liability to pull off the story.
The main trade off in the first stage is between the efficiency gain of tailored investment and information rent. Similar trade-off underlies the second stage game.
The idea and timing are clear. Wait for AO to detail it. To save time, I need to give him more guidance.