There is a new AI tool in your inbox every single week.
Image generators. Automation platforms. Agentic systems that promise to run your entire business while you sleep. And somewhere in all of that noise, you opened one of those tools, got confused in about four minutes, closed the tab, and went back to doing what you were already doing.
I get it. Because here is the real problem: the way AI gets taught to loan officers has nothing to do with closing loans. It is taught by people who are excited about AI. Not people who are in the mortgage business today, running files, training new LOs, and trying to figure out which of the 47 things on their plate actually moves the needle.
Most of what you need AI to do is not complicated. It is the simple stuff you already do, just slower, with more friction, and without the depth you could actually have if you had a little more time and a little better prep. That is what AI fixes. Not the hard stuff. The ordinary stuff you are already doing, done faster and done better.
The Framework
So let me give you the framework I use. And then I will give you two specific things to do this week.
One chat. One job. One output.
That is it. That is the whole thing.
The reason most LOs feel lost with AI is they open a blank chat and treat it like a search engine. They type something vague and general and get something vague and general back. And then they think AI is not that useful. But you would not hire a new team member, drop them in a chair, and say, "Hey, help me with stuff." You would say, here is the role, here is what I need from you, here is what good looks like, here are the files you need to do this well. AI is the same.
The input is yours. The output is AI's. The action is yours again.
Input, output, action. That is the loop. And the part most people get wrong is the input. Not because they are bad at writing prompts, but because nobody ever taught them to think about it like hiring and training a person. When I started getting better at AI, it was not because I got smarter. It was because I started building better inputs by asking AI to help me improve my own instructions. You read that right. I asked Claude if my prompt was any good. It said no. I asked why. I fixed it. The output got dramatically better. I have been doing that ever since.
You do not need to be a prompt engineer. You just need to give the thing a real job with real instructions, the same way you would onboard anyone new.
Tactic One: The Throwaway Chat
Before you call your account executive on a new program, before you try to pitch it to an agent, before you do anything with it, open a new chat in Claude or ChatGPT. Throw in the investor's program guidelines. Could be a PDF, a matrix, a page of product details, whatever they gave you. Then ask it three things.
What are the five things buyers most commonly misunderstand about this program? What questions will an agent ask me that I need to be ready for? And what is the minimum viable borrower profile so I know immediately whether someone qualifies?
That is it. You are not building anything. You are not creating an assistant. It is a throwaway chat with a specific job and a specific output. Ten minutes of prep. And you walk into that AE call or that agent conversation already sounding like you have been doing this program for two years instead of two days.
I watched an LO I coach go from total confusion on a DPA program to having a clear picture of three products, the qualifying criteria for each one, and a borrower profile she could pre-screen against, all in one throwaway chat. She would have spent three weeks trying to figure that out by reading the guidelines and waiting on email responses from the AE. She did it in an afternoon.
That is AI for loan officers. Not the image generator. Not the 12-step automation. The research and prep engine that compresses three weeks of learning into 45 minutes.
Tactic Two: The Scenario Bot
You probably have at least one investor right now that offers a DPA program. Maybe two or three. And if I asked you to tell me, off the top of your head, exactly which program a specific borrower qualifies for based on their income, purchase price, and credit score, you would probably have to dig. Maybe spend 20 minutes cross-referencing a matrix. Maybe send an email and wait.
Here is what I told my LO instead.
Take that investor's DPA matrix. The whole thing. Drop it in Claude inside a project. Give the project one job: you are a scenario bot. When I give you a borrower's profile, you tell me which program they qualify for, what the grant amount is, and what comp I can expect. That is the whole job. Nothing else.
She built that in ten minutes. Now every time she is on a call with a first-time buyer who needs help with the down payment, she takes what they told her, drops it in the bot, gets a clean answer in seconds, and sounds like the most prepared person that buyer has talked to in their entire home search.
That is a simple idea. Most DPA programs are not complicated once you know them. AI did not invent a new program. It just made the existing one faster to access and easier to apply.
The Point
AI is not going to close your loans for you. It is not going to replace the relationship, the urgency, the 7 a.m. call when the appraisal came in low and your buyer is panicking. But it will make you so much faster and so much sharper at the prep and research that every conversation you have with an agent or a borrower reflects that.
The LO who does three weeks of research in 45 minutes has a real advantage. Not because they are smarter. Because they stopped treating AI like a gimmick and started treating it like the research assistant they never could afford to hire.
Start with the throwaway chat. Pick one program. Drop in the guidelines. Ask those three questions. See what you get.
In part two, we take everything you just prepped and turn it into a program agents have never been offered before. That is where the real coffee meeting comes from.
