Who actually produces value when AI hits a brokerage floor

20 August 2026

Every conversation about AI in freight starts in the same place, which is whether it takes people's jobs. That is the wrong question, and asking it keeps getting in the way of the real one. The real one is what happens to a floor when the work that used to make everybody look the same stops taking any time.

Think about what a brokerage day is actually made of. Somebody builds a list. Somebody writes the follow-up that should have gone out on Thursday. Somebody chases a check call, then chases it again. Somebody tries to remember what a shipper said about their peak season on a call six weeks ago, gives up, and asks again in a way that makes it obvious they are asking again. None of that is skill. It is the tax you pay to get to the parts that are.

Here is the thing about that tax. It costs everybody roughly the same. Your best rep and your most average rep both spend the same twenty minutes rebuilding a prospect list, the same ten minutes writing a follow-up, the same afternoon chasing drivers. So on a Friday report, the difference between them is compressed. It shows up eventually, in the annual number, but week to week the friction acts like a leveller. Everyone is busy. Everyone looks like they are working. And if you have ever tried to explain to an owner why one rep who is clearly better only produces thirty percent more, that leveller is most of your answer.

Take the tax away and the picture changes fast, in a direction that surprises people.

The rep who was already good does not get thirty percent better. They get several times more done, because the thing holding them back was never their ability to have the conversation. It was that they only got eleven conversations a week and four of those were with accounts they should never have called. Give that person a list that is actually fit for the capacity you can cover, follow-up that writes itself, and a record of every promise anybody made on their accounts, and you find out what they were capable of the whole time. That is not a small lift. On a real floor it is the difference between a good producer and someone you have to build a retention plan around.

And the rep who was coasting does not get several times better. They get a bit faster at coasting. All the friction that was making them look busy is gone, and what is left is the part they were never doing much of. That is uncomfortable, and it should be, but notice what actually happened there. Nothing was taken from them. No system replaced them. The work they were not doing simply became visible, because the noise around it stopped.

This is why the framing of AI replacing brokers is not just wrong, it is backwards in a way that costs money. If you go into this looking to cut headcount, you will cut before you can see anything, which means you will cut on the basis of the same compressed picture you always had. The whole return here is that the picture stops being compressed. Cutting first throws that away and keeps the guesswork.

What actually happens on a floor that does this well is less dramatic and more useful. You stop needing to have an opinion about who your producers are. You can see it. Coverage gets faster because the person covering is not also doing data entry. The accounts you should be defending get defended, because somebody noticed the volume slipping in week two rather than in the quarterly review. And the conversations your managers have change character, from chasing activity to talking about the two or three deals that actually matter this month.

There is a version of this that goes badly, and it is worth naming. Drop tooling onto a floor and change nothing about how people are managed, and you get the same output with a new subscription line. The systems do not create urgency and they do not have the conversation with the rep who is not producing. They remove the excuse that the day was full. Somebody still has to act on what is now visible, and if nobody will, buy nothing, because you will have paid for a mirror you refuse to look in.

Twelve years in this business and one thing has not changed: the gap between the best person on a floor and the average one is enormous, and most of the systems we built quietly obscured it. Spreadsheets obscured it. CRMs that were really just databases obscured it. Activity metrics obscured it worst of all, because they measured the tax and called it work.

The interesting thing about this moment is not that machines can write an email. It is that the last few layers of that obscurity are coming off. Your A players are about to look like A players by a margin that is impossible to argue with, and everyone else is about to have a very clear number attached to them.

That is not a threat to good brokers. It is the best thing that has happened to them in a long time.