I will discuss a framework that approximately extends the classical theory of mechanism design for linear agents (following Myerson, 1981) to broad families of non-linear agent preferences. Optimal mechanism design for non-linear agents is generally analytically intractable. On the other hand, the presented framework shows that intuitions from linear agents approximately extend to non-linear agents. The talk will motivate the approach and analysis framework and sketch the main ideas that go into proving such approximation results. Joint work with Yiding Feng and Yingkai Li.