You'd send the fuel needed for the return trip ahead of the manned mission and leave it in orbit around Mars. You might even include the fuel needed for the landing, which would further reduce the mass you needed to decelerate into Mars orbit, which would multiply the savings because you wouldn't need to accelerate that fuel (orbit insertion and landing) to fly the transMars trajectory.

The fuel tanker would be unmanned, so it wouldn't need any life support system. It'd only need a minimal electrical system, guidance, attitude control and TT&C, propulsion, and some thermal. It might even be possible to use aerocapture for orbital insertion or at least to circularize the orbit following insertion like they did on MGS. The propellants would probably need to be hypergolic because they're storable. It isn't easy to store cyrogenic propellants in space for long periods of time. You could send the tanker well in advance of the manned mission so you'd be confident it arrived safely.

All of this becomes important because the greater the mass required at the beginning of the mission, the more it costs to start the mission. At several thousand dollars per kilogram to Earth orbit, saving a lot of money. Mass snowballs rapidly, and aerospace vehicles tend to cost by the mass.