Easy.
Don't.
Resonate with a cap. That's what everyone does, and with very good reason!
The fundamental problem is the very poor coupling between coils. Which means reactance dominates over real power transfer, by a factor (Q factor) of... 2, 5, 10 or more -- roughly, the inverse of the coupling factor.
With judicious use of ferrites (pole pieces and shielding), you can get a Q factor under 1, but only when things are very closely spaced, and the alignment must be pretty close as well. Basically, you get something like a split-bobbin transformer in that case, and any air gap or misalignment counts as air gap between windings -- at expense of coupling. Works for something like a cordless phone or toothbrush in a cradle, not so much for random sized items on a charging pad.
Resonance reduces the control problem to a matter of frequency, amplitude and/or phase control. Whereas direct drive simply draws whatever current the reactance does (again, because reactance is dominant), resonant can draw much more power (from the inverter), because load resistance acts to dampen resonance, and its absence therefore requires a control strategy.
In exchange, you have a lot of time to solve the control problem: rather than the per-cycle control that a lot of switchmode controllers employ, you can spend several cycles working out what frequency to choose, or amplitude to modulate to, and merely a peak current fault (latched) will do to protect the inverter. This is because bandwidth is also determined by Q factor, and a high Q means low bandwidth, means current changes less, from cycle to cycle.
Tim