I recently spoke to our agents about artificial intelligence at our general meeting. The idea that got the most hands in the air had nothing to do with tools or prompts. It came from neuroscience: cognitive load.

The plainest definition I have heard is: cognitive load is how much work your brain actually does. Handing that work to AI is like going from a 20-pound weight to a 2-pound weight. You are not nearly as strong.

The calculator is the counterpoint. Punching numbers into one probably means less brain activity, but it makes people more accurate and frees their minds for other work. Lighter weights only make you weaker if you stop lifting anything heavy. For an experienced agent, the listing description, the follow-up email and the first pass at a comparative market analysis (CMA) are reps that stopped building anything years ago. An agent who spends less energy on routine work can spend more on judgment, and judgment is what clients pay for. I teach agents to do this every week and I stand by it.

The same conversation had a second half, though, and it’s important.

Children. For a child, the struggle is how the brain gets built. It matures by wrestling with problems and learning from its own mistakes, and learning something new is hard precisely because it is new. A child who uses AI to skip the work skips the brain-building along with it. So the prescription flips. Adults should lower their cognitive load. Children need theirs raised, and they need to do that part without AI.

Somewhere in the middle of explaining that to a room full of agents, I realized I was describing the new registrant.

A first-year agent is obviously not a child. But in this business they are in the same position, because they have not yet built the thing the rest of us are offloading. A veteran who asks AI to draft a clause is outsourcing the typing. They have read hundreds of clauses and can spot a bad one before they reach the end of the first line. A rookie who asks AI to draft the same clause is outsourcing the part where they would have learned what a good one looks like.

Early research points the same way. In a small MIT study released last year, people wrote essays with ChatGPT, with a search engine or with nothing at all. More than eight in 10 of the ChatGPT group could not correctly quote a line from their own essay minutes after writing it. The work got done. Nothing stuck.

Now picture that rookie three years in. They have closed deals. Every CMA, every schedule and every difficult email to a co-operating agent went through a chatbot first. Then one day the chatbot is confidently wrong, as it often will be, and there is nobody at the desk who can tell.

That is the rule I keep coming back to. You cannot delegate work you would not recognize as done badly. AI makes a remarkable assistant and a terrible substitute for the experience you need to supervise it. And no regulator in this country is going to accept “the software wrote it” from the registrant whose name is on the offer.

None of this is an argument for keeping AI away from new agents. In fact, the opposite can be argued. The best way to teach anyone is one-on-one with a good teacher who knows them, and that has never scaled because it is expensive. An AI set up to teach could come close. The difference is who does the thinking. AI that makes the rookie do the work may be the best training tool this industry has ever been handed. AI that does the work for them produces an agent with a decade of closings and very little judgment to show for it.

So what does that look like inside a brokerage? These are the habits I would put in front of every mentor, trainer and manager working with new agents:

Hand first, then AI. The rookie writes the first CMAs, the first clauses and the first hard emails on their own. Then they ask AI to critique them the way a senior agent would, and they do the rewrite themselves.

Make it the examiner. Instead of asking AI to summarize the forms, the rules or the local market, new agents ask it to quiz them on all three until they stop getting answers wrong.

Explain it back. When a new agent brings you something AI helped produce, ask them to walk you through why each piece is there. If they cannot, it is not ready for a client.

Watch what you model. Children will model what you do. Rookies watch their mentors the same way. If the senior agents in your office paste and send without reading, that is the habit you are teaching.

The veterans in this business earned the right to lighten the load. They earned it by carrying that load for years. Our job, as the people training whoever comes next, is to make sure they get to carry it too.