What's Inside
How Realtors Work — Your full production workflow broken down by business area, with AI opportunities mapped to each stage. 🗺️
This Week in AI — Always-on AI assistants went mainstream this week, and the tools behind them got faster and cheaper. 🗞️
Top Picks — Six tools covering 24/7 agents, a faster Claude, small-business automation, voice, video editing, and a change coming to Gemini. 🔧
Deep Dives:
🤖 ChatGPT Dots — What OpenAI's always-on agent can do, what it costs, and whether it's worth the Pro upgrade
⚡ Claude Sonnet 5.5 — Why Anthropic's faster mid-tier model is the new default for everyday writing
🏪 Muse for Small Business — Meta's approval-first agent that works inside your business apps
🎙️ ElevenLabs v4 — Natural voiceovers for listing videos in 90+ languages
🎬 Vidmoat — A browser-based video editor you can direct with plain English
💎 Gemini Skills — Gemini Gems are being replaced in November, and what to do now
Mini Skill — Turn a long home inspection report into a clear, calm summary your buyer will actually read. ✍️
How Realtors Work: An AI Opportunity Map
Our monthly reference. The rest of the newsletter plugs into these workflows, so it's worth a skim even if you've been in the business 20 years.
1. Lead Generation
Sphere outreach, expired listings, FSBOs, online leads, door knocking, and social media all feed the top of the funnel. The work is repetitive by nature: pulling lists, researching owners, writing first-touch scripts, and posting consistently. Time loss usually comes from list prep and personalization, which is exactly where AI drafting and research tools save the most hours.
2. Lead Nurturing
Most deals come from leads that took months to convert, which means CRM follow-up, email sequences, text and call cadences, and regular market updates. The friction is consistency: follow-up slips when you get busy with active clients. AI helps by drafting personalized check-ins, summarizing where each lead left off, and keeping the cadence running when you're buried.
3. Buyer Representation
The buyer side runs from consultation through property search, showing coordination, offer writing, negotiation, and transaction management. Scheduling showings and chasing feedback eat hours, and buyers often need a lot of hand-holding to understand disclosures and reports. AI fits well in summarizing documents, drafting offer cover letters, and organizing showing logistics.
4. Seller Representation
Listing presentations, pricing strategy, prep and staging, photography coordination, marketing, open houses, offer review, and negotiation all land on your plate. Marketing assets (descriptions, social posts, video) and offer comparisons are where agents lose the most time. AI handles first drafts of listing copy, side-by-side offer summaries, and video editing far faster than doing it by hand.
5. Marketing and Content
Social posts, video, email newsletters, community content, and personal branding are what keep you top of mind. The problem isn't ideas, it's volume and consistency. AI is strongest here: content calendars, captions, video edits, and repurposing one piece of content into five.
6. Administrative and Operations
Contract management, deadline tracking, vendor coordination, commission tracking, and compliance are unglamorous but unforgiving. Missed dates and lost documents cost deals. Always-on agents and automation tools are starting to handle reminders, inbox triage, and status updates, though anything involving contracts still needs your eyes.
7. Client Experience and Retention
Closing gifts, post-close follow-up, referral cultivation, and review requests are what turn one deal into five. This is the area most agents neglect because there's no deadline forcing it. AI can draft anniversary notes, home-value updates, and review requests on a schedule so past clients don't go cold.
8. Professional Development and Recruiting
If you lead a team or plan to, this covers training, onboarding, and recruiting. Building scripts, training docs, and onboarding checklists takes time most producers don't have. AI is useful for turning your own know-how into repeatable playbooks.
This Week in AI — What It Means for Realtors
This was the week the "AI assistant that works while you sleep" stopped being a demo. OpenAI, Meta, Manus, and a handful of startups all pushed always-on agents: assistants with their own computer, connected to your email, calendar, and apps, that keep working after you close the laptop. The pitch is simple. You state the outcome, the agent figures out the steps, and it checks with you before anything sensitive goes out. For an agent whose day is a pile of follow-ups, reminders, and status updates, that's the most relevant AI shift we've seen this year.
The other theme was price. The best models got faster and cheaper to run, with Anthropic and OpenAI both shipping mid-tier models that come close to their flagships for a fraction of the cost. That matters even if you never look at a pricing page, because cheaper models are what make these background assistants affordable enough to run all day. Google also announced a powerful new model, Gemini 4 Argon, but it's limited to vetted security teams for now, so there's nothing for you to try yet.
One caution worth your attention: a company called Tavus showed an AI video-call avatar that nearly half of testers mistook for a real person. Real estate is already a top target for wire fraud, and live video was one of the ways clients could verify who they were talking to. It's a good week to remind buyers and sellers that wiring instructions only get confirmed by phone, at a number they already have, never one sent in an email or text.
Top Picks This Week
ChatGPT Dots
What it is: OpenAI's new always-on agent that runs on its own cloud computer and works across 4,000+ connected apps.
Realtor use case: A daily brief of today's showings, deadlines, and unanswered leads waiting for you at 8:30 AM.
Why it matters: It's the first major attempt at a true AI assistant that keeps working without you prompting it.
Claude Sonnet 5.5
What it is: Anthropic's new mid-tier model, about 30% faster than the last version and close to its top model on office tasks.
Realtor use case: Fast, high-quality drafts of listing descriptions, client emails, and offer summaries.
Why it matters: Near-flagship quality at a lower cost means better results on the plans most agents already pay for.
Muse for Small Business
What it is: Meta's personal agent, now with connectors for QuickBooks, Stripe, Shopify, Klaviyo, and more.
Realtor use case: Background help with bookkeeping and marketing tasks that holds every post or payment for your approval.
Why it matters: Approval-first design is the right model for anyone nervous about AI acting on their behalf.
ElevenLabs v4
What it is: ElevenLabs' newest voice model, rated the best speech generator available by a wide margin.
Realtor use case: Professional voiceovers for listing videos and reels, including in your clients' first language.
Why it matters: A narrated listing video no longer requires a studio or a voice actor.
Vidmoat
What it is: A browser-based AI video editor that takes plain-English instructions and edits until the job is done.
Realtor use case: Cutting dead air, adding captions, and resizing a walkthrough for Reels, TikTok, and YouTube Shorts.
Why it matters: It removes the editing step that stops most agents from posting video consistently.
Gemini Skills (replacing Gems)
What it is: Google is retiring custom Gemini Gems and migrating them to "Skills" starting in November.
Realtor use case: Any Gem you built for listing copy or client emails becomes a reusable Skill you can stack in any chat.
Why it matters: If you rely on Gems, you'll want to back up your instructions before the switch.
Deep Dives
🤖 ChatGPT Dots
Dots is OpenAI's headline launch from last week's DevDay. Think of a Dot as a personal assistant that lives in your ChatGPT sidebar on desktop, web, and mobile. You can message it or call it, give it a name, and it can message you first when something needs attention. Behind the scenes it runs on OpenAI's top model and its own cloud computer, so your laptop doesn't need to stay on. It connects to 4,000+ apps, and it can reply inside Slack or Microsoft Teams.
For a Realtor, the obvious first job is a morning brief. Connect your calendar and email, and ask it for a weekday summary of today's showings, contract deadlines, and leads you haven't answered. From there, it can monitor things for you, like flagging when a lender or title rep replies. The Rundown's setup guide recommends starting with local-computer access turned off, setting a custom rule so it asks before acting, and keeping sending, publishing, and purchasing out of its job entirely until you trust it.
The catch is price and access. Your first Dot is included only with ChatGPT Pro (now $100/month and up) and Business Premium plans, with wider access promised later. Reviewers liked the speed and quality of its research reports, but the desktop app asks for a lot of permissions by default, and you should skip most of them. Test it on low-stakes tasks first. One smooth answer doesn't tell you how it handles a changed closing date.
⚡ Claude Sonnet 5.5
Anthropic released Sonnet 5.5, the middle model in its lineup, and the reviews are strong. It runs about 30% faster than Sonnet 5, uses fewer steps to finish tasks, and nearly matches Anthropic's top model, Opus 5.5, on office-work tests. Pricing for developers didn't change, and Anthropic says most jobs cost up to 30% less because the model works more efficiently.
Why should you care? Most of what a Realtor asks AI to do (rewrite this email, draft this listing description, summarize this disclosure) doesn't need the most powerful model. A faster mid-tier model gets you the same quality draft in less time, and on a paid plan it stretches your usage limits further. Ben's Bites suggested that $20-plan users try moving some of their usual requests from Opus to Sonnet 5.5 with extended thinking turned on.
To try it, open Claude and pick Sonnet 5.5 from the model menu next to the send button. Run the same prompt you normally use on both models and compare. Model availability varies by plan, so check what your account shows. Keep using Opus for heavier work like comparing multiple offers line by line or reviewing long contracts.
🏪 Muse for Small Business
Meta's Muse is its answer to Dots: an always-on agent with its own secure computer and browser. Last week Meta added small-business skills, with connectors for Shopify, Stripe, QuickBooks, Klaviyo, and more. The key design choice is that Muse holds any post, payment, or outgoing message until you approve it.
That approval step is what makes this interesting for solo agents and small teams. You could have Muse reconcile expenses in QuickBooks, prep a marketing email, or monitor something on a schedule, and nothing goes out without a tap from you. The Rundown's guide shows a useful example: asking Muse to check local marketplace listings daily and log comparable items to a spreadsheet. Swap in something real-estate-flavored and you have a lightweight monitoring assistant.
Setup is a desktop app (Mac at launch) plus the phone app. Start with read-only permissions on every connector, including Gmail, and only raise access once you trust the output. Meta hasn't been specific about pricing for the business tier in what we read, so check before committing. It also connects to Meta's ecosystem, which is a plus if your marketing runs on Facebook and Instagram and a consideration if you're cautious about data sharing.
🎙️ ElevenLabs v4
ElevenLabs released Eleven v4, its newest voice model, alongside a faster Turbo version built for real-time use. Ben's Bites called it the best speech-generating model available "by quite a margin." It handles expressive speech in 90+ languages, supports multiple speakers in one clip, and lets you direct the performance (calm, upbeat, conversational).
For Realtors, the use is listing video narration. Write a 45-second script, pick a voice, and you have a polished voiceover for a property tour or reel without recording yourself. The multilingual support is the bigger opportunity: you can produce the same listing video in Spanish, Mandarin, or Vietnamese if that's who's buying in your market.
ElevenLabs has historically offered a free tier for testing, with paid plans for commercial use and higher volume, so confirm the current terms before using it in paid marketing. Two cautions. Only clone a voice you have permission to use, including your own. And review every script for fair housing language before it gets a voice, since describing neighborhoods or ideal buyers is where agents get into trouble.
🎬 Vidmoat
Vidmoat is a browser-based AI video editor featured in Future Tools this week. You tell it what you want in plain English, it plans the edits, and it keeps correcting until the task is done. Core features include cutting dead air, generating synced captions, removing backgrounds, and exporting for TikTok, Reels, YouTube, and Shorts.
This targets the exact bottleneck for agent video: you shoot a walkthrough or a market update on your phone, then never post it because editing takes an hour. With Vidmoat, you upload the clip and type something like "remove pauses, add captions, export a vertical version under 60 seconds." That's a realistic 10-minute workflow instead of an evening project.
There's a free tier and paid plans. Since it runs in the browser, there's nothing to install. It's a newer tool, so expect some rough edges, and always watch the final cut before posting. AI captions occasionally mishear street names and numbers, and a wrong price or address in a caption is a bad look.
💎 Gemini Skills (replacing Gems)
Google announced it's retiring Gemini Gems, the custom assistants many people built for repeat tasks, in favor of "Skills." Starting in November, personal-account Gems will migrate automatically to Skills. Unlike Gems, which you had to open separately, Skills are reusable instructions Gemini can apply automatically or stack together inside any chat.
If you built a "Listing Description Writer" or "Client Email Polisher" Gem, this is mostly good news. Instead of switching to a separate Gem, you'll be able to call your listing-copy instructions and your brand-voice instructions together in one conversation. That's a cleaner workflow for anyone who uses Gemini through Google Workspace.
What to do now: open each Gem you rely on and copy its instructions into a Google Doc. Migrations usually go fine, but a backup costs you two minutes. After the switch in November, test each Skill on a familiar task to confirm the output still matches what you're used to.
This Week's AI Mini Skill — 10 Minutes, Real Results
Turn a Long Inspection Report Into a Buyer-Friendly Summary
Inspection reports run 40 to 80 pages, and buyers tend to either panic over every note or skim and miss what matters. This skill turns the report into a clear, organized summary you can walk through with your client. It works on the free tier of Claude, ChatGPT, or Gemini, as long as you can upload a PDF.
Step 1: Remove personal details (1 minute) If your brokerage has rules about uploading client documents to AI tools, follow them. At minimum, don't add the client's name or financial details to the prompt. The property address is fine to leave out too.
Step 2: Upload the report and paste this prompt (2 minutes)
I'm a real estate agent. I've attached a home inspection report for a property my buyer client is purchasing.
Before you start, restate in one or two sentences what you think my goal is. Then wait for me to confirm.
Once I confirm, create a summary for my buyer with these sections:
1. SAFETY ISSUES – anything that could cause injury or is a code/safety concern (electrical, gas, structural, radon, mold, etc.)
2. MAJOR REPAIRS – big-ticket items likely to cost more than $1,000 or affect the roof, foundation, HVAC, plumbing, or electrical systems
3. MINOR REPAIRS – routine fixes a homeowner would handle over time
4. MAINTENANCE NOTES – items that are normal wear or simple upkeep, not defects
5. QUESTIONS TO ASK – items where the inspector recommended further evaluation by a specialist
Rules:
- Use plain language a first-time buyer would understand.
- Include the page number from the report for every item.
- Do not estimate repair costs unless the report itself gives one.
- Do not invent anything that isn't in the report. If something is unclear, say so.
- Keep a calm, factual tone. No alarming language.Step 3: Confirm the restatement (1 minute) The AI will tell you what it thinks you're asking for. If it's right, reply "Confirmed, go ahead." If it's off, correct it now. This one step prevents most bad outputs.
Step 4: Check the output against the report (5 minutes) Spot-check the Safety and Major Repairs sections using the page numbers. AI can occasionally misread a table or miss an item, and this is the section that drives your repair request. If anything looks wrong, tell it: "Item 3 is on page 22, not 14. Recheck it."
Step 5: Use it (1 minute) Paste the summary into an email to your buyer or bring it to your review call. It also makes a great starting point for your repair request, since everything is already sorted by priority.
What not to do: Don't present this as a replacement for the inspection report, and don't give your opinion on repair costs or whether the house is "a good buy." The summary helps your client understand the report. The decisions stay with them and their inspector.
The AI Edge for Realtors is published weekly. If you found this useful, forward it to another agent in your office.