What's Inside
This Week in AI - Big models keep leapfrogging each other, but the useful news for agents is smaller: AI that builds working tools, watches your pipeline, and screens your texts. 🗞️
Top Picks - Five tools and workflows you can try this week, from calculators to scam screening. 🔧
Deep Dives:
🧮 ChatGPT Intelligent UI - Turn one answer into a net sheet or payment calculator you can adjust live
📊 Claude Dashboards - See which deals need attention without building a spreadsheet
🗓️ Scheduled AI Follow-Up Drafts - A Monday-night task that preps your emails before you wake up
🖼️ Nano Banana 2.1 in Google Flow - One listing graphic becomes a matching set of social images
🛡️ Gemini Scam-Text Screening - Opt-in protection for texts from numbers you don't know
Mini Skill - Have a second AI attack your listing price pitch before the seller does. ✍️
This Week in AI: What It Means for Realtors
The headlines were all about the model race. Meta shipped a frontier model called Muse Wonder, Google made Gemini 4 Argon available to everyone, and Anthropic's Opus 5.5 still sits near the top of the rankings. You don't need to pick a winner. What matters is that the best models are now close enough in quality that the tool you already use is probably good enough. The real gains come from what you build with it.
That's the pattern in this week's practical news. ChatGPT can now turn an answer into an interactive calculator. Claude can turn connected data into a live dashboard. Scheduled AI tasks are showing up in real small-business workflows, like one owner who has an agent draft his late-invoice reminders every Monday night and saves about an hour a week. These are the kinds of jobs that eat Realtor time: follow-ups, tracking, and explaining numbers to clients.
One more thing to keep in your back pocket. Reporting this week showed advertisers can target people inside ChatGPT based on what they're asking about, and the personalization setting reduces ads but doesn't fully turn them off. Treat any AI chat like a search box. Keep client names, addresses, and financial details out of prompts unless your brokerage has approved the tool.
Top Picks This Week
ChatGPT Intelligent UI
What it is: A new ChatGPT feature that turns an answer into a small working tool with editable inputs.
Realtor use case: Build a seller net sheet or buyer monthly payment calculator you can adjust live in front of a client.
Why it matters: You stop pasting static numbers and start showing clients how changes affect the result.
Claude Dashboards
What it is: A Claude feature that builds a live dashboard from a data source you connect.
Realtor use case: Pull your pipeline from a spreadsheet and see which leads and deals are overdue.
Why it matters: You get a Monday-morning view of what needs attention without building one by hand.
Scheduled AI Follow-Up Drafts
What it is: A recurring AI task that runs on a schedule, checks a list, and leaves drafts for you to review.
Realtor use case: Every Monday night, draft check-in emails to leads who haven't heard from you in 30 days.
Why it matters: The drafting is done before you start your week, and you still approve every send.
Nano Banana 2.1 in Google Flow
What it is: Google's image model, which can turn one product photo into a matching set of campaign images.
Realtor use case: Turn one listing or "just sold" graphic into square, portrait, and wide versions for different platforms.
Why it matters: Consistent branding across platforms takes minutes instead of an hour in a design tool.
Gemini Scam-Text Screening
What it is: An opt-in Android Messages mode where Gemini reads chats from unknown numbers to judge whether they're scams.
Realtor use case: Fewer fake showing requests, wrong-number hooks, and lookalike "closing" texts getting through.
Why it matters: Agents get texts from strangers all day, and reply-based scams are the kind keyword filters miss.
Deep Dives
ChatGPT Intelligent UI
Intelligent UI lets ChatGPT build a small interactive tool from your prompt instead of just a block of text. In The Rundown's example, you ask for a workshop cost calculator with editable cost, budget, and attendance fields, then change the headcount and watch the total and budget gap update. The same idea works for a seller net sheet, a buyer cash-to-close estimate, or a rent vs. buy comparison.
For a Realtor, the win is the client conversation. Instead of saying "it depends," you change the sale price or the commission and show the result on screen. Setup is minimal: open ChatGPT in a browser, start a chat with GPT-6, and describe the inputs and the rules that should stay fixed. Plan on 10 to 15 minutes for your first one.
The caveat is accuracy. Test it before it goes in front of a client. Enter a case you can calculate by hand, then try an edge case like zero or a very high price. Tell the AI which numbers are inputs and which rules never change. And treat the output as an estimate, not a lender or title company figure. I couldn't confirm from the newsletter which plan tiers include the feature, so check your account.
Claude Dashboards
Claude's new Dashboards feature builds a live report from data you connect. The Rundown's walkthrough uses a sales pipeline: open deals, total value, and overdue deals, with filters by owner. Swap "deals" for "leads and active transactions" and you have a view of your business that updates when your source does.
The benefit lands in lead nurturing and transaction tracking, where things slip because nothing is in one place. The guide calls for a paid Claude account and a data source you approve. Start with a small test spreadsheet, keep access read-only, and spend about 30 minutes on it.
Check the numbers before you trust them. The guide has you open a figure's query, confirm which rows count, and test whether a new record shows up after refresh. A chart can look current while running on old data. Also, don't connect a full client database until your brokerage is comfortable with it.
Scheduled AI Follow-Up Drafts
Two Rundown reader workflows point at the same idea. One owner has Claude Cowork run every Monday night, check a late-invoice list against a Google Sheet, and put reminder emails in his Drafts folder, with a summary report. He says it saves about an hour each week. The Rundown's own CEO runs a Friday scheduled Claude task that prepares a review report before he wakes up.
For a Realtor, swap the invoice list for your sphere or lead list. Keep a sheet with name, last contact date, and a note. The scheduled task flags anyone past your threshold and drafts a short, personal check-in using your template. You open Drafts Monday morning, edit, and send.
Setup is the real cost. Expect an hour or two to get the sheet, template, and schedule working, and it needs a tool that supports scheduled tasks and email access. The drafts-only rule matters: don't let any AI send client emails without your review. Both examples are reader-reported, so test with a few contacts first.
Nano Banana 2.1 in Google Flow
The Rundown's guide uses Google's Nano Banana 2.1 model in Google Flow to turn one product photo into a matching campaign set: a square image first, then portrait and wide versions that keep the same product, light, and style. Then you add a short headline and check every word.
For a Realtor, the product is a listing photo, a headshot, or a "just sold" announcement. Make one approved version, then generate the formats you need for Instagram, Facebook, and LinkedIn. Flow shows a usage cost before generating, so look at it first.
The caveat is honesty. Anything that alters a property's appearance, including AI staging or cleanup, needs to be disclosed per your MLS rules and state requirements, so check your local rules. The guide also recommends keeping a text-free copy and adding exact brand fonts and your logo in your usual design tool, since AI text rendering can be off.
Gemini Scam-Text Screening
Google is adding an Enhanced Protection mode to Android's Messages app. When a number not in your contacts texts you, cloud-based Gemini can read the conversation to judge whether it's a scam. A second layer checks unfamiliar links on-device, but that part needs a Pixel 10 or newer, a Galaxy S26, or select devices.
This matters for agents because you answer strangers for a living. Reply-based scams, like wrong-number and fake-inquiry texts, adapt to each victim and have no link for a filter to catch. It's opt-in and rolling out over the coming weeks, so look for it in Messages settings.
The caveats are real. It only covers unknown senders, and scammers often move people to WhatsApp or Telegram. Google hasn't shared how often it flags real texts by mistake or misses scams. Google says messages stay private, and The Neuron noted it couldn't verify that. It's not an answer to wire-fraud emails, so keep verifying payment instructions by phone with a known number.
This Week's AI Mini Skill: 10 Minutes, Real Results
Have a second AI attack your listing price pitch
The Neuron highlighted a technique Anthropic uses internally: point an AI at finished work with the job of attacking it, not approving it. Here's the Realtor version. Before a listing presentation, have a different AI poke holes in your pricing recommendation. You'll walk in already knowing where the seller will push back.
What you need: your pricing rationale (a few paragraphs or bullets), and two free AI chats. Use the same tool you wrote it in for step 1, and a different one (ChatGPT, Claude, or Gemini) for step 2.
Write your pricing pitch in plain bullets: suggested list price, 3 to 5 comps and why they're comparable, days on market expectations, and your marketing plan. Remove the seller's name and address.
Open a different AI chat than the one you used to write it.
Paste the prompt below, fill in the brackets, then paste your bullets underneath.
Read the top problems. Mark each one as "real flaw," "likely objection," or "ignore."
Ask this follow-up: "For the three most likely objections, give me a calm 2-sentence response I'd say out loud."
You are a skeptical seller who thinks their home is worth more than my recommended price. Below is my pricing recommendation for a [type of home] in [neighborhood/city], written for a seller who wants to [sell fast / get top dollar].
1. List at most 7 objections or weak points, ranked by how likely a seller is to raise them times how much they could hurt my credibility.
2. For each, say what would have to be true for it to matter and how likely that is in a normal market.
3. Separate real flaws in my reasoning from hypothetical risks, and flag anything I should ignore.
4. Do not rewrite my pitch yet.
My recommendation:
[paste your bullets here]What to do with the output: Fix the real flaws in your comps or numbers, and rehearse the three objection responses. Check any market stats the AI cites against your MLS data, since it can't see your market. Ranking by likelihood and cost matters: a reviewer told to find problems will list every possible one, and the ranking shows you which to ignore.
The AI Edge for Realtors is published weekly. If you found this useful, forward it to another agent in your office.
Questions or tool suggestions? Reply to this email.