Every AI story that matters to agents.

Turn sales call notes into follow-up emails

A Rundown guide walks through pasting call notes into an AI model, asking for a plan recommendation and short unsent follow-up for each lead, then checking each against the actual offer. It stresses that a requested discount is not an approved one and measures time to a usable reply.

Why it matters: Post-showing and post-call follow-up is where deals are won, and AI can draft it in seconds as long as you check every promise against the facts.

Try this: After your next three showings, dictate quick notes and have AI draft a personal follow-up for each, then edit and send within the hour.

Agent Image's SiteAI turns agent websites into answer engines

Agent Image launched SiteAI, an AI layer for agent websites that answers visitor questions with branded, magazine-style pages built from the agent's content, IDX feed and LiveBy neighborhood data. It replaces the contact form with a lead brief showing budget, neighborhoods, intent and a suggested opening line, with fair housing guardrails built in.

Why it matters: Consumers increasingly expect to ask questions instead of clicking menus. A site that answers and pre-qualifies can make your website a real lead source again.

Try this: Ask ChatGPT or Claude for the 10 questions buyers in your market ask most, then check whether your website clearly answers each one.

OpenAI's GPT-6.1 Terra: faster, cheaper, fewer errors

OpenAI released GPT-6.1 Terra for ChatGPT Plus, Pro, Enterprise and Edu users and its Codex tool, built for computer use and speed at about half the API price of GPT-5.6 Terra. OpenAI says the factual error rate on its trick-question test fell from 53.6% to 16.7%.

Why it matters: A faster, more reliable mid-tier model makes ChatGPT's agent mode more practical for routine web chores. Fewer hallucinations is still not zero, so facts need checking.

Try this: Give ChatGPT's agent mode one real chore, like gathering permit history or HOA details for a listing from public sites, and verify every fact it returns.

Reality check: Fewer errors isn't zero errors. OpenAI's own test still showed a 16.7% error rate on trick questions, so every fact it gathers about a property gets checked before it reaches a client.

Staffer used ChatGPT renders to brief a contractor

A Rundown staffer and his wife sketched a bedroom layout, had ChatGPT render it into a floor plan, refined wood, finishes and views by voice, and took the final images to their contractor. Inspiration came from Google Images and Pinterest.

Why it matters: Buyers who say 'this house needs too much work' often just can't picture it. A quick concept render can keep a dated listing in the running.

Try this: Upload a photo of a dated kitchen and ask ChatGPT for a renovated concept image; share it with buyers clearly labeled as a concept, never as a listing photo.

Reality check: A renovation render is a concept, not the house. Label it 'concept' every time, or it can read as misrepresenting the property.

Reader's scheduled Claude agent updates a living doc weekly

A Rundown reader set a scheduled Claude agent to refresh a living workout log every Friday, pulling from an API, a Google Drive folder and the document itself, with no custom code. It sends a notification every run, even when nothing changed.

Why it matters: The same pattern can produce a weekly seller update with showings, feedback and market stats that runs without you remembering to do it.

Try this: Draft a weekly seller update template, then set a scheduled AI task to fill it from your showing feedback notes every Friday for your review.

Anthropic files to go public as ANT

Anthropic publicly filed its S-1 to list on Nasdaq under the ticker ANT, reporting $16.3B in revenue for the first half of 2026. The filing also warns it is unclear when an AI agent's actions legally bind the organization that deploys it.

Why it matters: Not a tool to use today, but it signals Claude is a long-term business platform worth learning. The 'who is liable for an agent's actions' question applies to your business too.

China's Kimi K3.1 Pro nears the AI frontier

Moonshot released Kimi K3.1 Pro, which ranks No. 4 on Artificial Analysis's intelligence index, with open weights due Nov. 4. It is priced well below top US models.

Why it matters: Cheaper powerful models will lower the cost of AI features inside real estate software. Be cautious about where client data goes in apps built on foreign models.

Give past clients something worth keeping

The Tide Report describes how Lunden Group built a 45+ page local business directory for current and past clients and followed it with an off-season specials spreadsheet. It cites a Psychological Science study finding people underestimate how much recipients value thank-you letters.

Why it matters: One useful touch a month keeps you top of mind for referrals without asking for anything. AI makes building these local guides fast.

Try this: Ask Claude or ChatGPT to draft a categorized local vendor directory for your town, then verify every listing and add your own picks before sending.

The Agency builds its own private AI platform

The Agency is building an in-house AI platform that unifies MLS, transaction, financial and marketing data so its 4,000+ agents can query it in plain language. Sensitive client and transaction data stays on self-hosted open-source models while commercial models from OpenAI and Anthropic handle tasks like market analysis.

Why it matters: Large brokerages are choosing to keep client data off public AI, which raises the bar for how independents handle privacy. Expect recruiting pitches built on 'our AI knows our data.'

Try this: Split your AI tasks into public-data work (market summaries, social posts) and client-data work, and keep the second group in a business plan with model training turned off.

Comments and DMs beat Instagram likes

The Tide Report cites ManyChat's estimate that an Instagram like is worth about 7.5 cents and argues comments and DMs drive business, not likes. It suggests comment-to-DM prompts and tracking DMs, leads and appointments instead of likes.

Why it matters: Automated comment-to-DM flows turn a listing post into a conversation, which is what actually leads to showings.

Try this: On your next listing post, ask people to comment one word (like 'TOUR') and send the details by DM, then track how many turn into conversations.

AI voicemail assistant answers your missed calls

Hola AI Voicemail answers missed calls, talks with callers, and sends summaries by push, email or SMS, while filtering spam into a searchable inbox. It works with your existing number through call forwarding and is a paid app.

Why it matters: Missed calls during showings are missed leads. A voicemail assistant that summarizes and texts you the gist helps you call back the right people first.

Try this: Forward your missed calls to an AI voicemail tool for one week and track how many callbacks you prioritize differently.

New startup tracks how AI chatbots recommend agents

AuthoritySignal AI, a Las Vegas startup, measures how agents, teams and brokerages show up in AI-generated recommendations and identifies gaps to fix. Former REMAX president and T3 Sixty executive Nick Bailey joined as a founding member.

Why it matters: 'Who is the best agent in my town?' is becoming a chatbot question, and the answer draws on your reviews, bio and published content. AI visibility is the new local SEO.

Try this: Ask ChatGPT, Claude, Gemini and Perplexity 'Who are the top real estate agents in [your town]?' and record whether you appear and why.

Reality check: No service can guarantee that ChatGPT recommends you. Reviews, a clear bio and local content help, and anyone promising a #1 spot is selling hope.

Canva launches Craft, its own AI image model

Canva launched Craft, its first in-house image model, trained on designer-made templates and built to plan a layout before drawing so results convert into editable designs. It debuted at No. 4 on DesignArena and is live for all Canva users and via API.

Why it matters: Most agents already build flyers and social posts in Canva. A model that understands layout and text placement means fewer broken AI graphics and faster just-listed posts.

Try this: Use Craft to generate a 'Just Listed' graphic from your listing details, then edit the text layers instead of starting from a blank template.

Reality check: AI-made graphics can still get text wrong. Never let it type the price, square footage or your license number. Lock those fields in a template.

Where AI belongs in listing marketing, and where it doesn't

A HousingWire column argues AI should handle captions, photo cleanup and virtual staging, while prices, unit numbers, license numbers and MLS notices belong in fixed templates mapped to listing data. Fully AI-generated flyers can get exact facts wrong and drift off brand.

Why it matters: A wrong price or missing license number on an AI flyer is a compliance problem, not a typo. Locked templates plus AI-written copy give you speed without that risk.

Try this: Build one Canva flyer template with locked fields for price, beds and baths and your license info, and use AI only for the headline and description.

1Password lets AI agents log in without seeing passwords

1Password for Agents lets AI assistants sign in to sites with the user's approval without ever seeing the password, rolling out first for Meta's Muse and Grok Bot, with Claude Duet, ChatGPT and Town planned. It is included at no extra charge for Individual and Family accounts.

Why it matters: Agents juggle MLS, showing, CRM and portal logins. Never paste passwords into a chatbot; approval-based sign-in is the safe way to let AI work inside those sites.

Try this: Follow The Rundown's 1Password guide on a low-risk test account: deny one request, approve one, then revoke access and sign out.

If it's copy-paste, let AI do it through connectors

Rundown educator Nate Grahek argues that any copy-and-paste step between apps should be handled by an AI connector, and suggests a 'read, then write, then send' ladder for trusting AI with your inbox, drive and calendar. He recommends having AI write drafts while you still hit send.

Why it matters: Agents lose hours moving info between email, CRM, docs and MLS. Connectors with a human-sends rule are a low-risk way to win that time back.

Try this: Connect your email to ChatGPT, Claude or Gemini in read-only mode and ask it to list every client email from this week still waiting on a reply.

Arizona indie brokerage runs on homegrown AI

Scottsdale's Incyte Realty built Console, a proprietary CRM, CMA and marketing platform, plus 'Morgan,' an AI assistant that handles reception, sign calls, texts and scheduling. The broker says average production rose from 6.3 to nearly eight transactions per agent, a self-reported figure.

Why it matters: Small independents can now build tooling that rivals national brands. An AI receptionist on sign calls is a realistic first step for any team.

Try this: Point your yard-sign number at an AI answering or voicemail service and review the transcripts weekly for missed buyers.

Let Town prepare your next meeting brief

A Rundown guide shows how Town, a business-focused AI assistant, turns email and calendar details into a meeting brief, catches an unconfirmed time change and saves the process as a reusable routine. It recommends keeping changes behind approval.

Why it matters: Walking into a listing appointment knowing every prior email, date and open question is a competitive edge.

Try this: Before your next listing appointment, have an AI assistant summarize every email with that seller into a one-page brief.

Make.com tracker flags past donors who haven't given

A Rundown reader built a Make.com workflow that reads platform emails, logs donations in a spreadsheet, and compares against prior years to flag past donors who have not given yet. He can ask it questions by email, like who gave over $1,000 before but not this year.

Why it matters: Swap donors for past clients and you get a sphere tracker that flags who you haven't touched this year.

Try this: Export your past-client list and ask AI to flag everyone you have not contacted in 12 months, sorted by past sale price.

Critical PHP flaw puts 70% of websites at risk

PHP pushed emergency patches for two critical flaws, dubbed PHP2Shell, that let attackers take over servers without logging in; OpenAI's models helped researchers find them. Amazon reported active exploitation, and the FBI and UK NCSC advise taking unpatched systems offline.

Why it matters: WordPress, which powers many agent and team websites, runs on PHP. A hijacked site can leak lead data or serve malware under your name.

Try this: Email your website host or IDX vendor today and ask in writing whether your site's PHP version is patched for PHP2Shell.

Anthropic launches Claude Duet, an always-on agent

Anthropic introduced Claude Duet, a proactive agent app built on Opus 5.5 that sends a morning brief, monitors connected apps through the day and flags what needs attention. It launched as a research preview on Android, iPhone and web, with a waitlist for Claude Pro and Max subscribers.

Why it matters: An assistant that watches your inbox and calendar is the admin layer most solo agents lack. Early access lets you test it before your competitors do.

Try this: If you are on Claude Pro or Max, join the Duet waitlist and list the three daily checks (new leads, showing confirmations, contract deadlines) you would hand it first.

Reality check: An always-on agent that reads your inbox sees everything in it, including client finances. Grant it the fewest connections it needs, and keep sending in your own hands.

Gemini now customizable with skills and 40+ connected apps

Google added a central 'customize' hub to the Gemini app with custom skills for repeat tasks and a consistent tone, plus connections to 40+ Google and outside services. It is available to all Google Workspace customers and personal Google accounts, with most connected apps in English in the US.

Why it matters: Skills let you teach Gemini your voice and formats once instead of re-explaining in every chat. That is how you get listing copy and follow-ups that sound like you.

Try this: Create a Gemini skill called 'My listing voice' with three of your best past descriptions and your fair housing do-not-use word list.

Gemini comes to Google Maps on Apple CarPlay

Google brought conversational Gemini to Google Maps on CarPlay for US drivers, handling multi-step requests like finding a place and routing there while avoiding tolls. Drivers can also ask about stops and report traffic hands-free.

Why it matters: Agents spend hours driving between showings. Hands-free multi-step requests are safer and faster than tapping the screen.

Try this: On your next showing day, plan stops entirely by voice ('find coffee near my 2 p.m. showing, then route me there') and note what it gets right.

Real Brokerage launches Leo Pro pay-per-task AI agents

Real unveiled Leo Pro, a library of 36+ AI agents inside its reZEN platform that create listing videos, pursue cold leads and move transactions forward, asking for approval before important actions. It is priced in tokens (a monthly plan with 20,000 tokens, $9.99 intro price) instead of per seat, and is only for Real agents.

Why it matters: Pay-for-completed-work pricing is a preview of how brokerage AI will be sold, and it sets a benchmark for what agents at any firm should expect from their brokerage's tools.

Try this: Ask your broker which agentic tools (lead follow-up, listing video, transaction reminders) are on the roadmap and how they will be priced.

Reality check: Pay-per-task AI still needs your approval on anything client-facing. An AI agent that follows up with a cold lead is still you texting that lead, under your license and the consent rules.

Gemini adds a hub for small business owners

Google added a small business hub to the Gemini app that connects to Google Business Profile, Google Ads, Workspace, Canva, Dropbox and others, saves a brand style for AI-generated visuals, and turns spreadsheets and receipts into financial charts. Several connections are limited by account type and region.

Why it matters: A solo agent is a small business. Pairing your Google Business Profile with on-brand visuals and expense analysis in one app saves real time.

Try this: Save your brand colors and logo as a brand style in Gemini, generate three social graphics, and compare them with your current templates.

ChatGPT now lets you lock private chats

ChatGPT added Locked Chats, which hide individual conversations behind a PIN, Face ID or Touch ID and give them separate memory. It does not add encryption or change how OpenAI stores data, and it is rolling out across Free, Go, Plus, Pro, Business, Enterprise and Edu plans.

Why it matters: Agents hand their phones to clients and colleagues constantly. Locking chats about negotiations or commissions adds privacy, but it is not a vault for client data.

Try this: Lock any ChatGPT chats that contain negotiation strategy or commission talk, and review which client details you have pasted in before.

Perplexity Computer now runs fully on your Mac

Perplexity's Computer agent can now run fully locally on Apple silicon Macs for Pro, Max and Enterprise subscribers, working across Notes, Mail, Messages and other apps without sending personal data to the cloud. Perplexity recommends at least 36GB of memory for local use.

Why it matters: Running an agent on your own machine is one answer to the 'don't paste client data into AI' problem for agents with sensitive files.

A 78-year-old runs his life by voice in a Claude Project

A Rundown reader who manages a homestead and a senior center built 'Taskmaster,' a plain-text ledger in a Claude Project that he updates by speaking into the Claude app. Claude tracks partial progress and, when asked 'What's next?', weighs deadlines, weather and energy.

Why it matters: Agents capture details in the car between showings. Voice notes into a running project log is a simple system that doesn't need a new app.

Try this: Create a Claude or ChatGPT project called 'Active deals,' and after each showing dictate notes and ask 'What's next for this client?'

Zillow puts home search inside Claude

Zillow launched a Claude connector that lets people search for-sale and rental listings, check home values and estimate affordability inside a chat. In the same conversation users can request a tour, connect with an agent or explore financing, adding to Zillow's existing ChatGPT app and Gemini rentals experience.

Why it matters: Buyers are starting searches inside chatbots, and Zillow is positioning itself as the default handoff to an agent there. If you are not visible where the search starts, Zillow decides who gets the lead.

Try this: Connect Zillow in Claude (Customize > Connectors), search your farm area, and note how listings, broker attribution and the 'connect with an agent' step appear.

Reality check: Zillow, not you, decides which agent gets the 'connect' tap inside the chat. Today there's no way to opt in or pay to be that agent, so treat this as a visibility warning, not a lead source.

Survey: agents pay for ChatGPT, not real estate AI

An Inman and Lofty survey of 460 respondents found 62% of agents pay for a general-purpose AI assistant while only 14% pay for a real-estate-specific tool. 41% of agents caught a factual AI error before it reached a client, and 17% of leaders reported a disclosure, fair housing or MLS compliance issue tied to AI material.

Why it matters: Your general assistant is doing most of the work, so the edge is how well you prompt it and how carefully you check it. Revenue gains are still rare, with only 4% reporting more appointments.

Try this: Audit your AI subscriptions: cancel any niche tool you have not opened in 30 days and put that money toward a paid general assistant plan.

Reality check: Paying for ChatGPT doesn't mean it's working: only 4% of agents in the survey reported more appointments. The edge comes from a few repeatable workflows, not from the subscription.

Turn a content spreadsheet into a working app

A Rundown guide uses Airtable Omni to turn a content plan into an app showing what needs review, who owns it and what is overdue, then adds an automation that moves finished drafts into review. It warns to leave missing owners blank rather than writing 'Unassigned.'

Why it matters: Agents posting regularly need a simple content calendar. Turning a spreadsheet into a lightweight app keeps posting consistent.

Try this: Put next month's social post ideas into Airtable and ask Omni to build a review board with due dates.

Feds ban DeepSeek, including tools built on it

The Department of Homeland Security ordered civilian federal agencies to remove DeepSeek products and third-party services built on its models, citing concerns about sensitive data reaching China. Agencies have 60 days to request exceptions.

Why it matters: The ban covers tools that don't carry the DeepSeek name, a reminder that you should know which AI model sits under the apps you put client data into.

Try this: Ask your CRM and transaction-management vendors which AI models power their features and where your data is processed.

Economist: AI's housing impact runs through jobs

First American's Odeta Kushi says AI's labor impact is still uncertain, with entry-level workers feeling some pressure, and notes productivity gains take a long time to show in data. She links a 'low hire, low fire' job market to fewer job moves, which dampens buying and selling.

Why it matters: Buyer demand tracks job mobility. Watching how AI reshapes local hiring helps explain your market to clients.

Anthropic data: AI is shrinking junior hiring

Anthropic published a study of its own workforce finding junior roles fell from 27% to 9% of technical hires and employees' AI agents logged six times the hours of the people running them. Agents sat idle about 80% of the time waiting on a human to respond.

Why it matters: Teams are hiring fewer junior roles because AI handles first drafts, and the human reviewer is the bottleneck. For agents building a team, the next hire may be a TC plus AI rather than another assistant.

Try this: Time how long you take to review and send one AI-drafted email, then set a goal to cut that review time in half with a saved template.

BoldTrail and Real Geeks now talk to your AI assistant

Inside Real Estate launched a BoldTrail MCP so AI assistants can review priorities, log notes, schedule follow-ups and flag leads with no recent outreach. Real Geeks launched AI Connect, letting agents use their chatbot of choice to update website copy, pages, blog posts and design.

Why it matters: CRMs are becoming something you talk to through ChatGPT or Claude instead of logging into. 'Who haven't I called?' becomes a 10-second question.

Try this: If you use BoldTrail, connect it to your AI assistant and ask: 'Which leads have not heard from me in 30 days?'

Reality check: Connecting your CRM to an AI assistant gives that assistant your clients' contact details and notes. Start read-only, and check your brokerage's data policy before you connect.

AI voice agents now call your cold leads

CINC launched CINC Call AI, powered by Flair Labs, to call new and high-intent leads, ask about timeline and motivation, and live-transfer ready leads to an agent. Virginia's Yeatman Group uses Mod AI to call very cold leads, though some leads show 'AI fatigue' and decline to engage.

Why it matters: Speed-to-lead is being automated, which raises the bar on every agent's response time. Calling and consent rules still apply, so clear any AI caller with your broker first.

Try this: Pull 20 leads older than six months and text each one personally this week, then compare your reply rate with what an AI calling vendor promises.

Reality check: AI callers and texting are covered by the TCPA and the do-not-call rules. Consent and opt-out still apply, and some leads are already tuning out 'AI fatigue' calls.

Turn a ChatGPT answer into a working calculator

A Rundown guide shows how to use ChatGPT's Intelligent UI to build an interactive cost calculator with editable inputs, then test edge cases and add a budget buffer. It advises telling ChatGPT which inputs change and which rules stay fixed.

Why it matters: A simple seller net sheet or buyer cash-to-close calculator you can adjust live in a meeting is a strong client tool.

Try this: Ask ChatGPT to build a seller net sheet calculator with editable price, commission, payoff and closing costs, and test it against a recent closing statement.

Rowan Cheung's Friday AI 'factory inspection'

Rundown CEO Rowan Cheung runs a scheduled Claude task every Friday that picks random areas of his company to inspect, gathers samples and data into a Notion report, and checks Slack to confirm last week's fixes. He shared the exact prompt in The Rundown's Workflow Hub.

Why it matters: Agents rarely audit their own business. A weekly automated review of a few random leads, listings and follow-ups catches dropped balls before clients notice.

Try this: Schedule a weekly AI task that pulls three random leads and one active listing and reports what follow-up is overdue.

ATTOM adds AI-ready context to property data

ATTOM introduced Semantic Views, which add business context to its property data delivered through Snowflake so AI systems can better interpret how the data connects. ATTOM calls it the first stage of an ongoing effort.

Why it matters: Property data vendors are rebuilding for AI, which means better automated valuations and market reports inside the tools agents use.

MLSs race to set AI rules for listing data

Northstar MLS, CRMLS and partners are building Project NexusRE to track and control how AI systems access and use listing data, while Northwest MLS requires AI tools to use a licensed data feed. CMLS is drafting best practices for agents using MLS data in AI tools.

Why it matters: How you feed MLS data into ChatGPT or Claude may soon be governed by MLS rules. Dropping unlicensed exports into AI tools could become a compliance issue.

Try this: Read your MLS's current rules on exporting listing data to third-party tools before building any AI workflow on MLS exports.

ChatGPT can now read and act across your Chrome tabs

OpenAI upgraded ChatGPT's Chrome integration so it can compare information across open tabs and, in Work mode, fill forms, draft emails and organize tabs. OpenAI also launched GPT-Live-1.1 with 13 new voices and better non-English support.

Why it matters: Comparing listings, lender quotes or vendor sites across tabs is daily agent work, and ChatGPT can now build that comparison for you.

Try this: Open five competing listings in Chrome and ask ChatGPT for a table of price, price per square foot, days on market and standout features.

Reality check: Comparing listings across tabs is fine. Copying MLS data into a chatbot may break your MLS rules, so check them before building a workflow around it.

Reader used ChatGPT to audit a vet bill line by line

A Rundown reader photographed an itemized veterinary estimate, uploaded it to ChatGPT with the doctor's notes, and asked which charges were likely unnecessary. ChatGPT flagged four of 24, and the vet agreed two could be dropped.

Why it matters: The same move works on contractor quotes and inspection repair estimates, helping buyers and sellers ask better questions before agreeing to a number.

Try this: Upload a repair estimate from a current deal and ask ChatGPT which line items to question and what to ask the contractor; have the contractor confirm before relying on it.

Kling 4.0 makes 30-second AI video clips

Kuaishou launched Kling 4.0, which generates up to 30 seconds of video in one pass with smoother motion and better continuity through complex camera moves. Its Omni Reference feature accepts up to 15 reference assets and 10 keyframes.

Why it matters: Longer, steadier clips make AI b-roll and neighborhood intros usable for Reels. Never use it to show a property in a way that misrepresents it.

Try this: Generate a 30-second lifestyle intro (local coffee shop, trail, downtown) for your next market video, and keep the listing footage real.

Amazon's Bee wearables record and summarize conversations

Amazon's Bee launched the Frontier wearable ($74) and the Eclipse phone attachment ($99), which capture conversations and calls and turn them into notes, summaries and reminders. Amazon says audio is deleted after processing and Eclipse asks the other party's permission before recording calls.

Why it matters: Auto-summarized showings and client calls are tempting, but recording people raises consent issues, and recording laws vary by state.

Krea 3 brings precise, cheap AI image edits

Krea released Krea 3, an image editor that changes only a selected region, takes up to six reference images and outputs 2K to 4K, at about $0.03 to $0.06 per image. It can be fine-tuned on a specific person or style for consistent results.

Why it matters: Region-only edits are handy for headshots and marketing graphics. Any edit that changes how a property looks must be disclosed or avoided.

Try this: Use Krea 3 to make a consistent set of branded headshot variations for your social profiles.

States sue OpenAI over ChatGPT safety claims

At least eight states, including Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska and West Virginia, sued OpenAI, alleging ChatGPT harms children and that the company misled users about safety testing. One suit also cites AI agents that hacked other companies.

Why it matters: Regulators are now actively going after AI companies, and consumer-facing AI that gives advice is in the spotlight. Real estate regulators could follow with licensee rules.

When AI makes a mistake, the agent pays

Compliance consultant Summer Goralik, a former California Department of Real Estate investigator, says agents, not AI tools, carry the license and legal risk when AI output is wrong. She urges brokerages to set and actually enforce AI policies and expects an early enforcement action to define what regulators expect.

Why it matters: Every AI-written listing remark, text or chatbot answer goes out under your license. A policy you signed but nobody enforces will not protect you.

Try this: Write a one-page personal AI rule: what you never paste into AI, what always gets a human read before sending, and what a website chatbot may not answer.

Facebook lets users tell its algorithm what they want

Meta launched 'Your Algorithm' on Facebook, letting users describe in plain language which topics they want more or less of, with recommendations adjusting in real time. It follows similar controls on Instagram and Threads.

Why it matters: If people can ask for 'more local real estate and home tips,' clearly topical, local content gets an edge, and vague posts become easier to filter out.

Try this: Rewrite your next three Facebook posts so the topic and town are obvious in the first line, for example 'Asheville market update: October.'

Meta's Muse Wonder joins the frontier

Meta released its Muse Wonder models, with Muse Wonder Full ranking No. 2 on Artificial Analysis's intelligence index behind Claude Opus 5.5. Wonder is available through Meta's API and subscriptions, and Meta also released Muse Spark Open as a free open-weight model.

Why it matters: Meta's models increasingly power the ad tools and assistants inside Facebook, Instagram and WhatsApp, where many agents market. Better models there mean better automated ad creative and messaging.

Ryan Serhant launches a daily-posting AI avatar

SERHANT founder Ryan Serhant partnered with HeyGen to launch @HeySerhant, an AI avatar based on his likeness and voice that will post daily real estate, leadership and market content for 12 months. Serhant approves topics and messaging, and the avatar answers common questions.

Why it matters: AI clones of well-known agents are here, which raises questions about disclosure and trust. Used carefully, an avatar can scale FAQ-style video without filming every day.

Try this: Record a 30-second FAQ answer with an AI avatar tool, label it clearly as AI, and compare engagement with a video you filmed yourself.

Reality check: An AI avatar of you still speaks for you. Disclose that it's AI, approve every script, and know that some clients will trust it less.

eXp lets agents plug any AI into its systems

At eXpcon, eXp announced 'MCP First, AI of Choice,' which lets agents connect Claude, ChatGPT or another assistant to their production numbers, leads, transactions, training, CRM records and Canva. It is the first piece of AGNT OS, a broader platform due in early 2027.

Why it matters: The bring-your-own-AI model means the assistant you learn now follows you across tools and brokerages. Picking one and going deep pays off.

Try this: Choose one assistant (ChatGPT, Claude or Gemini) as your daily driver so its memory, skills and connectors build up in one place.

Claude adds live dashboards; Docs and Slides go free

Anthropic launched Claude Dashboards (beta, paid plans), which build auto-updating dashboards from connected data, and Claude Motion (Team and Enterprise), which turns ideas into short animated explainers exported as MP4. Claude Docs, Slides and Design also came out of beta on every plan, including Free.

Why it matters: A free AI slide builder covers listing presentations, and Motion can turn a market report into a short social video.

Try this: Use Claude Slides to draft a six-slide listing presentation from your CMA notes, then edit it to your brand.