Plenty of real estate teams use AI chat every day and still haven't taken real work off anyone's plate. Someone asks ChatGPT for a better listing description, someone else has it review a contract, and the TC still spends the day logging into one system after another, retyping the same details and chasing the same people.
Alberto Zesati has spent more than 20 years in real estate lending and now runs two businesses, a mortgage business and an AI agency, under the name The AI Lender. In both, people and AI work side by side every day. In the third episode of Keyed In, Alberto opened by naming who the session was for: the transaction coordinators, admins and operations teams who do the work nobody at the closing table sees.
"The agent gets the handshake at the closing table, but you are the one who gets the file to the finish line."
Alberto Zesati, The AI Lender
The session had two goals, making that work easier and making its value impossible to miss. Alberto's way in is to pick the task before you pick the tool. You score your recurring work with a simple audit, write the steps down, and build AI around the highest-scoring task that a person can still review. Alberto then showed three systems built that way, an SOP builder, a deal status dashboard and a value report. Alberto was clear that these are fictional examples prepared for the session, and that they cover the work before, around and after the core transaction workflow ListedKit handles, rather than replacing it.
Why do AI tools still leave so much manual work for a real estate team?
Alberto asked the audience how many systems they touch on a normal workday. The answers came in at four, ten and "10 plus," with one TC pointing out that working for agents at several brokerages means a different set of logins for each. Email, the CRM, the transaction management system, compliance software, contracts and a long tail of websites all hold part of every deal.
"It's all the handoffs between those systems."
Alberto Zesati, The AI Lender
Each time information moves from one place to another, somebody copies it, somebody chases an update, and sometimes something gets missed. That's why Alberto thinks the question people ask most, which tool is better, gets things backwards. ChatGPT, Claude and the rest are all good. What matters is what you point them at.
"The right question is what task in my business should I automate first?"
Alberto Zesati, The AI Lender
How do you decide which real estate task to hand to AI first?
Alberto's answer is a framework called the 7 R's Operations Audit. Take one task you do often and ask seven yes or no questions about it. Every yes is a point.
- Repeat: Do you do this task over and over, like opening escrow every time an offer is accepted?
- Retype: Are you copying information from one place to another, such as from a purchase contract into your CRM, calendar or email?
- Remind: Are you chasing people for signatures, documents, deposits or updates?
- Reply: Are you answering the same questions again and again, starting with "where's my deal?"
- Report: Are you building reports or sending status updates to agents or your team?
- Rules: Does the task follow the same steps every time?
- Review: Can a person quickly check the result before it goes out?
The higher the score out of seven, the stronger the candidate. Alberto added three rules before any AI gets to work. First, a score of five or more makes a task a strong candidate, and if two tasks tie, start with the one that takes the most time. Second, if a person can't quickly review the result, don't let AI handle it alone. AI can draft a client status update, but it shouldn't decide on its own that a contingency has been removed and tell the client.
"AI drafts and a person approves."
Alberto Zesati, The AI Lender
Third, if the process doesn't have clear steps, write them down first, because "AI can't follow a process that only exists in your head."
To show how it works, the group scored a task most TCs know well, chasing missing HOA documents. It repeats on every transaction with an HOA, it means retyping the HOA's contact details and request, it involves reminding people more than once, and agents keep asking whether the documents have arrived. It follows a repeatable process, and a person can review the request before it goes out. The only no was reporting, so it scored six out of seven, a strong candidate to automate at least part of the work.
How do you turn a process that lives in your head into an SOP?
That third rule is where most teams get stuck, so Alberto's first demo started there. The starting point is a voice memo. You, or the agent you support, talk through how a process actually runs, such as the first 48 hours after an accepted offer or bringing a new agent onto the team, and hand the recording or its transcript to AI.
"You don't have to explain it perfectly. You just have to get the steps out of your head."
Alberto Zesati, The AI Lender
In the demo, a rambling brain dump ("first thing I do is get a contract, there's open escrow, blah, blah, blah") became a numbered process: confirm the contract, build the deadlines, open escrow, send the intro email, and so on. A nine-step onboarding process got the same treatment, and then Alberto switched on a heat map that marks each step for AI, a virtual assistant or a person. Several steps could run with no person involved, while verifying the new agent's license stayed with a person.
Karan Khanna, ListedKit's co-founder, said the same thing comes up when setting up new ListedKit users. People often skip steps when describing their process because they think of them as "something I handle," and those are often the steps a tool could take over.
"Tell the AI everything. Let it tell you what it can automate."
Karan Khanna, ListedKit
Once the process is written down, the same SOP can become a flowchart, a checklist a new hire works through, and even a short training quiz to check that they understood it. If you're building out training for a new coordinator, our transaction coordinator training guide covers the skills side.
"A business without an SOP is a chaotic business."
Alberto Zesati, The AI Lender
How can agents and clients see where a deal stands without calling the TC?
With the process documented, Alberto turned to the question every TC hears most: where's my deal? The second demo was a Where's My Deal dashboard. A sample agent, Jordan, had four active deals, and two were flagged, one red and one yellow.
"In two seconds, we can identify right away where to focus our attention."
Alberto Zesati, The AI Lender
Opening the red card showed why: the earnest money deposit was due that day and escrow hadn't received it. The yellow card showed an appraisal that came in $6,000 under price, with the buyers needing to decide what to do by a set date. Neither needed a call to the TC to explain, because the reason sat right on the card.
The same dashboard can be built at three levels. Agents see their own pipeline, a team lead sees every deal across the team, and a client view shows a buyer or seller where their transaction stands, what comes next and whether anything is needed from them. Alberto suggested a TC could build the client view for each agent they support and send it out automatically every week, which improves the client's experience and makes the TC's work visible to the agent at the same time. Alberto noted that the data for a dashboard like this can come straight from ListedKit, which already tracks each deal's timeline and due dates. For the team-wide version inside ListedKit, see the team dashboard.
How can a transaction coordinator prove their value to agents?
The third demo was Alberto's favorite. A TC catches mistakes, chases deadlines, follows up again and again, and puts out small fires before they become big ones. Most of that happens where the agent never sees it.
"If the agent doesn't see it, they may not understand what your worth is."
Alberto Zesati, The AI Lender
The Value Report makes that work visible. Using fictional sample numbers, Alberto showed how it estimates what an agent gets back for every dollar they spend on TC services, and how that figure grows when the agent's time is worth more. Moving the agent's hourly value from $150 to $300 took the sample return from about $2 back per dollar to about $3. The report also lists the files closed and, for each one, the problems caught along the way, such as a wrong phone number on one sample file and an unsigned lead-based paint disclosure on another, along with the tasks completed and the hours handed back to the agent.
The report can run monthly, quarterly or yearly, cover a whole team, and go out automatically, with a review step so the TC can check it before it's sent. Alberto's point was about timing as much as visibility. When business slows down, everyone looks at what to cut.
"It's gonna be pretty hard to cut your services as an expense."
Alberto Zesati, The AI Lender
Where should you start this week?
Alberto closed the main session with homework that takes about 15 minutes. Write down three tasks you do repeatedly, starting with the ones that drive you crazy. Run each through the 7 R's and give it a score. Then start with the highest scorer, as long as a person can review it and the steps are clear, and ask whether it's just for you or whether your whole team needs to be involved. You don't need to automate your entire business tomorrow. Get one task working, then move to the next.
Karan pointed out that the three demos answer requests that usually come in separately. An agent wants paperwork handled, a TC wants faster contract intake, and a team lead wants a dashboard. Those may be different jobs on a team, but they all start from the same place, the deal. Alberto compared it to building a house: the foundation is your context and your data, and the dashboards are the rooms you add on top.
From the Q&A
Can an AI assistant do a transaction coordinator's job? Alberto said an assistant can handle parts of the job, but Alberto wouldn't let one run a transaction alone and would add steps where a person checks in. Karan said the answer depends on volume. An agent doing one to three deals a year who has the time can use an AI tool to help, but in that case the agent is the TC, and the responsibility for keeping the deal moving stays with them. For a TC supporting a team or a high-producing agent doing five to seven deals a month, someone still needs to oversee what the AI sends, for example to stop it from emailing the seller details about the buyer that shouldn't go to them. As Karan put it, the role is "evolving, but not going away," and one TC can take on more deals than before. Our solutions page for transaction coordinators covers that side of the question.
Why does the same AI tool give some people better results? Context. Alberto said general-purpose apps built for everyone give everyone the same generic experience and make more mistakes. When you take the time to give Claude or ChatGPT a context file about you, your business, your team, your systems and where you want the business to be in six months, "the output is significantly different, because it's that foundational context layer."
Can AI work with the MLS? Yes, depending on the tool. Some MLSs have started offering connectors that give AI tools access to their data, and a browser extension can log in and navigate the MLS the way a person would. Alberto's advice was to give it an SOP rather than a loose request, because "it works good when you have an SOP built," and to build in a step where a person reviews and approves before anything is posted, since your license is on the line.
Key takeaways
- Score three tasks this week. Run each one through the 7 R's and start with the highest scorer, five out of seven or more.
- Keep a person on every review step. If no one can quickly check the result, AI shouldn't handle that task alone.
- Record one process as a voice memo. Turn it into a written SOP, then mark each step for AI, a virtual assistant or a person.
- Answer "where's my deal?" before anyone asks. Show agents and clients the status and the next step in one place.
- Show agents what you caught. Send a regular summary of the problems you prevented and the time you gave back.
For more from Keyed In, see Episode 2 on how directors of operations use AI to help their teams handle more deals, and Episode 1 on using AI on your real estate team without losing the human touch. Follow Alberto on Instagram at @theailender or on LinkedIn.