Yeah this is pretty much my assumption. I would imagine that the person doing the judging may not be on site as it's probably cheaper and easier to have a bigger workforce in one office, flex the people in shifts, load balance across all ports with the gates, etc.

I do wonder how much is automated. For example, is the gate fully automated in calculating a risk score and then referring you to a border agent if above a threshold? Is a person looking at the details in realtime to decide that, or are they just doing facial recognition, or are they only involved when the gates fail.