It is widely suspected that AGN activity ultimately sweeps galaxies clear of their gas .
We work out the observable properties required to achieve this .
Large–scale AGN–driven outflows should have kinetic luminosities \sim \eta { L _ { Edd } } / 2 \sim 0.05 { L _ { Edd } } and momentum rates \sim 20 { L _ { Edd } } / c , where { L _ { Edd } } is the Eddington luminosity of the central black hole and \eta \sim 0.1 its radiative accretion efficiency .
This creates an expanding two–phase medium in which molecular species coexist with hot gas , which can persist after the central AGN has switched off .
This picture predicts outflow velocities \sim 1000 - 1500 km s ^ { -1 } and mass outflow rates up to 4000 ~ { } { M _ { \odot } } { yr } ^ { -1 } on kpc scales , fixed mainly by the host galaxy velocity dispersion ( or equivalently black hole mass ) .
All these features agree with those of outflows observed in galaxies such as Mrk231 .
This strongly suggests that AGN activity is what sweeps galaxies clear of their gas on a dynamical timescale and makes them red and dead .
We suggest future observational tests of this picture .