My actual week with AI, job by job.
I sell homes with Compass in Danville, and I run my own practice on the same systems LeadJens builds for clients. This page is what runs each week, what each piece does, and where I still do the work myself. If a word is new, the glossary has a one-line answer. For the plain-English definition, read what an AI agent really is. To work on this together, start at AI for agents.
Most agents use AI as a writing tool, and many haven't felt a difference yet. In the National Association of REALTORS® 2025 technology survey, ChatGPT led the AI tools agents reported using (58%), and 46% of agents said AI hadn't made a noticeable impact on their business, while about half reported a positive one. The agents who do feel it use AI for repeat jobs with a clear trigger: who to follow up with today, which new listing fits which buyer, what to bring to a showing. That's how I use it, and the rest of this page is the list.
A follow-up runner tells me who to contact today and why. It reads my pipeline and my notes on each buyer and seller, then gives me up to five people, ranked by how close each one is to a decision. Every name comes with a suggested draft and the reason: an unanswered question, a home that fits, a market change, a promise I made, a task that's due or a reply waiting on me. "It's been a while" never counts as a reason. It suggests at most one touch per person every 14 days; a home that fits them, or a reply from them, overrides that. It never sends anything. The rules are spelled out in AI follow-up: what to automate and what to keep human.
A listing match checks new inventory against every active buyer. When coming-soon and off-market alerts arrive, it compares each home with what each buyer told me they need, checks the listing page itself, and flags only the strong fits. The weak maybes stay off my desk.
A broker tour preview turned the morning tour email into a plan. It read the day's broker tour list, cross-referenced my active buyers, and flagged which homes were worth the drive. It produced more than 100 dated tour plans between April and August 2026.
I built a showing brief for tour days; I used it with clients in July 2026. It's a print-ready packet: a card for each home with notes on how it lines up with what the client asked for, the questions worth asking at the door, and a few nearby stops.
Every request goes through one AI chief of staff first. I call mine Wags. Instead of a dozen separate chats, everything goes to one place that sorts it, pushes back when two commitments collide, and sends it to the right working file. It's the pattern I teach first, because every other system plugs into it.
Every Monday, a scheduled task builds my command sheet. It's a one-page PDF of what only I can do that week, laid out by day and built from my own records. It has run every Monday since early October 2026.
A post-close playbook runs the 30 days after the keys change hands. It covers the thank-you, when to ask for a review, the referral thank-yous, and a "Tale of the Sale" story about the home with no client named. That story is how one closing becomes a month of content, and it's covered in content that doesn't sound like AI.
AI never sends to my clients. It can draft a message; I review every one before it goes out. I also use an email and text sharpener on my own drafts: it checks what I wrote, tells me what's unclear, and tightens it. My house rules for any client message: the answer first, a real question last, emails at 150 words or fewer, texts at 300 characters or fewer, and never "just checking in."
Numbers follow once they're logged. I haven't logged hours per task yet, so I won't quote a figure I can't show you. For a benchmark, RPR's February 2026 survey of 225 agents found 68% of agents using AI save at least an hour a week. What I can tell you today is which jobs no longer depend on my memory: who's due for a call, which new listing fits whom, what to bring on tour, and what happens after a closing.
Questions agents ask about using AI
Start with one paid chat assistant, such as ChatGPT, Claude or Gemini, and one repeat job you already do every week, such as follow-up or tour prep. Use it on that job for a month before adding anything. Agents who buy a platform first tend to end up with features they never turn on. If you want a guided start, my free beginner's guide to Claude walks through the first setup.
Less than you'd think. If you can explain a task to a new assistant in plain English, you can explain it to AI. I'm not a developer by background; I built every system on this page from a standing start inside my own practice. The skill that matters most is knowing your own process well enough to describe it step by step.
Not today, in my experience. I use AI to plan tour days, track follow-up and prepare post-close steps, but nothing I run manages escrow deadlines on its own, and I keep it that way on purpose. Contract dates and documents stay with a person, whether that's you or a coordinator you trust.
AI is already doing the remembering and sorting in my business, and it does that well. It can't walk a buyer through a house, read a seller's hesitation or negotiate across a table, and that's the part clients hire an agent for. The bigger near-term risk is being invisible when a client asks ChatGPT who to hire. The free AI Visibility Score shows where you stand.
ChatGPT, Claude and Gemini all have free plans that are fine for trying a task or two. The paid individual plans add more use, better models and longer memory for documents, which matters once you rely on them daily. Either way, read the data settings before pasting anything about a client into any AI tool.
Jens Hansen is a REALTOR® with Compass in Danville, CA (DRE #02274665), and the real estate work described on this page is from his own practice.
Want this in your business?
A 15-minute call is enough to find the first job worth handing to AI in your week.