Any autoregressive process is necessarily invertible but a stationarity condition must be imposed to ensure uniqueness of model for a particular autocorrelation structure. A moving average process on the other hand is stationary but in order for there to be a unique model for a particular autocorrelation structure an invertbility condition must be imposed. This invertibility condition is usually built into a software to ensure uniqueness of model fitting. The root of the characteristic equation should lie outside the unit circle for such a moving average model.
A moving average model is a linear combination of innovations ei's. That is Xt = et+a1*et-1+a2*et-2+ ... + aq*et-q is a moving average model of order q (MA(q)). The current innovation is et and the et-i's are the past innovations. For invertibility of an MA(1), a1 < 1.
The following discussion in stackexchange should give you an intuitive explanation of the invertibility of moving average processes (see the comment by Rob Hyndman): https://stats.stackexchange.com/questions/50682/what-is-the-intuition-of-invertible-process-in-time-series
Invertibility refers to linear stationary process which behaves like infinite representation of autoregressive. In other word, this is the property that possessed by a moving average process. Invertibility solves non-uniqueness of autocorrelation function of moving average.