Observations have offered strong support for the theory of inflation, but many details of the physics of the early Universe are still unknown. An observation of primordial isocurvature modes would change our view of the early Universe and point towards multi-field inflation. In this talk, I will describe some recent work towards the calculation of second-order effects from isocurvature modes,focussing on the intrinsic bispectrum of the CMB. I will show the effects of general isocurvature modes to be small, but that a large enhancement is found for compensated isocurvature modes, given their large potential amplitude. The sensitivity of these second-order effects to this compensated mode allows us to severely constrain its amplitude using the non-observation of the predicted bispectrum.