The crawl economy flipped

Three data points tell the story.

The crawl has flipped. Alli AI’s January to March 2026 analysis of 24 million HTTP requests across 78,000 pages found that ChatGPT alone made 3.6x more requests than Googlebot. Add OpenAI’s separate training crawler, GPTBot, and the combined OpenAI footprint is 3.8x Google’s (Search Engine Journal, sponsored by Alli AI).

The clicks are collapsing. Ahrefs re-ran its AI Overviews study using December 2025 data and found the presence of an AI Overview correlates with a 58% lower click-through rate for the top-ranked page (Ahrefs). Semrush clickstream data shows 92 to 94% of Google AI Mode sessions end without a click to any external site (Semrush). Similarweb pegs the zero-click rate on AI Overview searches at roughly 83% (Clickvision zero-click analysis). The Semrush debate is worth acknowledging: its own same-keyword study argues AI Overviews do not cleanly cause zero-click behavior (Semrush AI Overviews study). But the direction of travel is not in dispute.

The demand has moved. McKinsey’s October 2025 research found that half of consumers now intentionally choose AI-powered search, a majority call it their top source for buying decisions, and adoption spans generations, including a majority of baby boomers. Roughly 50% of Google searches already carry AI summaries, projected to exceed 75% by 2028. McKinsey estimates unprepared brands face 20 to 50% declines in traditional search traffic (McKinsey). Gartner forecast a 25% drop in traditional search volume by 2026 back in early 2024 (Gartner). That forecast is arriving on schedule.

Cloudflare’s data explains the economics behind the shift. Googlebot crawls roughly 14 pages per referred visitor. Perplexity sits near 195. GPTBot is over 1,000. ClaudeBot crawls nearly 24,000 pages for every single referral (Cloudflare, Seomator analysis of Cloudflare Radar). Some engines cite sources and drive clicks. Others absorb the answer and keep the user. Either way, your visibility depends on one thing: whether you get named in the answer.

The small slice of AI-referred traffic that does land on a website converts dramatically better than traditional sources. Knotch reports LLM-referred visitors convert at roughly twice the rate of other traffic, in one-third the number of sessions, per data shared in Conductor’s 2026 benchmarks (Conductor 2026 AEO/GEO Benchmarks). Volume is small today. Quality is not.

Real estate discovery queries are exactly where this lands first. “Best agent in Greenwich.” “Is Darien a good place to buy in 2026?” “How competitive is the buyer market in The Berkshires right now?” These are research-stage questions. They have already moved.

A counterpoint is worth surfacing here. Conductor’s 2026 benchmarks (first-party data from a vendor that sells AEO platform services, so worth treating with appropriate skepticism) show real estate has the lowest AI Overview coverage of any industry they track, around 20% on average (Conductor AIO Volatility Analysis). The reason is structural. “Best agent in Greenwich” or “three-bedroom homes in Darien under $1.5M” do not generate generic AI Overviews. They pull users straight into ChatGPT and Perplexity instead. The discovery layer is migrating regardless. It is just migrating past Google’s overlay and into the answer engines themselves.

Being cited is the new ranking

The old structure was straightforward. Let the crawlers in. Rank the page. Receive the click. That contract is not dead, but it has been demoted. Perplexity, Google AI Overviews, and ChatGPT Search still drive traffic. They drive it to the source they cite, not the page they rank. Being cited is the new ranking.

Call it AEO, GEO, answer-layer engineering. The label does not matter. The work does. Engineer your entity presence so that when AI is asked who the best agent in a market is, it names you. That is the whole game.

No vendor solves this for you

My inbox is full of “AI visibility” pitches. Dashboards that monitor where your brand shows up in ChatGPT. “GEO services” that track citations. Alerts that notify you when Perplexity mentions you. They measure what they cannot produce. They show you the scoreboard. They do not play the game.

The work lives on your side of the wall. And here is the sleeper stat from McKinsey: brand websites supply only 5 to 10% of the content AI engines reference in generated answers. The rest comes from reviews, press, directories, portals, MLS feeds, and third-party profiles. You can own your website completely and still be a minority voice in your own AI answer. No vendor rewrites your entity graph for you at scale. No subscription service cleans your NAP data. No dashboard writes your neighborhood guides. The only path is work.

The DIY stack

None of this requires code. All of it requires work.

Start with table stakes. NAP consistency (Name, Address, Phone) across every directory, portal, and profile. AI engines reconcile entity data across sources, and every inconsistency is a reason to cite someone else instead of you. Google Business Profile, fully claimed and maintained, with current photos, monthly posts, and a real review system. Your brokerage profile, your personal website, and the major portals: Zillow, Realtor.com, Redfin, Homes.com. A meaningful share of AI answers about local professionals still traces back to GBP and portal data.

Then the sustained lever: content. Hyperlocal. Data-dense. Recent. Neighborhood guides with actual numbers, not platitudes. Monthly market reports. Buyer and seller question libraries with real answers, not filler. Answer-first structure on every page, where the first sentence answers the primary question directly and each section stands alone so AI can extract it cleanly. Most agents skip this layer. It is the one that compounds. Conductor’s data backs it up: blog content was the most cited page type in real estate AI Overviews by a wide margin, with over 15,000 pages cited versus roughly 8,000 for the next category. Multiply your surface area: YouTube with proper titles, descriptions, and transcripts. LinkedIn articles. Medium or Substack. Long-form, structured, hyperlocal content is what AI reaches for first.

Then the infrastructure. JSON-LD schema (RealEstateAgent, Person, FAQPage, Review, AggregateRating) applied consistently across bios, listings, and about pages. Entity consistency across every public surface. Review velocity as an ongoing system, not a once-a-year campaign. LLMS.txt is worth implementing as a low-cost hedge, but be honest: it is a proposed convention, and adoption by OpenAI, Anthropic, and Google is still unconfirmed.

Then the third-party surface area. The credibility layer AI weighs heavily. Google Reviews on a velocity system. Rate My Agent. FastExpert. HomeLight. RealTrends rankings if you have them. U.S. News Real Estate. Press mentions in local and trade outlets. Podcast appearances with searchable transcripts. MLS data hygiene. This is where most of the answer actually comes from.

And the layer almost everyone ignores: community signals. AI engines lean heavily on Reddit, Facebook Groups, Nextdoor, and local community forums for local-market queries. You cannot manufacture authentic presence in those communities. You can show up consistently, contribute real value, and earn the references over time. It is slow work. It is also where AI looks when it wants a real human voice instead of a brand pitch.

A markdown file, an AI copilot, and a few weekends. I have been doing this work myself. If I can, an agent can. If an agent can, a brokerage can. This is the Builder, Not Coder moment for real estate discovery.

The industry problem

This is not an agent problem. It is not a brokerage problem. It is both.

Agents waiting for the firm to fix visibility will stay invisible, because the firm cannot rewrite every agent’s bio or claim every agent’s directory profile. Firms treating this as agent marketing will leave the brokerage’s own entity data broken, and the answer layer will describe them using whatever Zillow, Redfin, and Reddit say.

Two layers, one system. The brokerage owns the platform and the entity foundation. The agent owns hyperlocal content and review velocity. Neither layer wins alone. And neither gets rescued by a tool.

The Conductor 2026 benchmarks tell you exactly where residential brokerage stands today. Top five domains AI cites for real estate queries: Hines, Public Storage, CBRE, ExtraSpace, Colliers. Top five brand mentions: same list plus Zillow. Not one residential brokerage. Not one agent. The entire residential industry is currently absent from the AI citation layer for its own queries. ChatGPT drives 95.2% of AI referrals in real estate, the highest concentration of any industry Conductor tracks. The opportunity is wide open. Whoever shows up first owns the surface.

Start this weekend

The seller in Westchester will open ChatGPT again next month. So will the buyer in Greenwich, the family in Darien, the empty nester in the Berkshires. The question they ask will be answered with or without you. Being named in that answer is not a purchase. It is work.

No vendor rewrites your entity graph. No dashboard writes your neighborhood guides. No subscription claims your GBP profile or cleans your NAP data. Pick one page. Rewrite it answer-first. Claim one directory you have been ignoring. Ship one neighborhood guide with actual numbers. Write the schema template once and apply it across your site.

You are not optimizing alone for Google anymore. You are engineering for the layer above it.

When that seller asks the question next month, your name is either in the answer or it is not. Nobody else decides that for you.

And for more than three years, a meaningful share of them have been using consumer AI tools, often with client data, almost always without any governance layer between their keyboard and a model running somewhere in the cloud.

Brokerages that have responded by purchasing enterprise AI have done something useful. But that alone is not enough.

Enterprise AI licensing is table stakes. Governance is the strategy. And in a contractor-based workforce, governance carries an obligation most broker-owners and leaders have not yet internalized. If you do not provide a sanctioned path, your agents will find an unsanctioned one. The absence of a solution exposes agents and their broker to liability and risk.

The six layers of a real governance stack

A serious brokerage AI posture has six layers. Each one depends on the others.

1. A written AI policy that defines acceptable use, prohibited use, and data handling expectations.

2. An operating model that assigns ownership, cadence, and enforcement.

3. Technical controls including SSO, admin access, audit logs, and data loss prevention.

4. An honest accounting of the BYOD (bring your own device) and independent contractor reality.

5. A plan for shadow AI, meaning the consumer tools your agents are already using.

6. Agent enablement through training, communication, and culture.

Most brokerages have one or two of these. Very few have all six. The layer most operators underweight is not the technical one. It is the structural reality those controls have to function inside.

Pressure point one: the independent contractor reality

Most business-focused Gen AI programs assume W2 employees on managed devices. That is not residential real estate.

A typical residential real estate firm has a smaller population of employees using corporate hardware and a larger population of independent contractors on personal devices. You can require federated SSO for the tools you provision. You cannot require it for a tool an agent installs on their own laptop with a personal credit card. You can offer training. You cannot mandate that an agent closes the consumer tab when the sanctioned one is open.

This is not a complaint. It is the operating reality every broker-owner and operator has to design around. The governance stack does not ignore the independent contractor status. It needs to be built on top of it.

Two consequences follow. First, enforcement is always partial. Technical controls create a sanctioned perimeter, not a total one. Second, the soft layers, meaning policy, training, and sanctioned tooling, carry more weight in brokerage than they would in a typical enterprise. In many cases they are the only levers that reach the agent at all.

Pressure point two: shadow AI and the consumer tool exposure

Shadow AI is endemic in residential brokerage, though the pattern is not unique to this business. It is not a story about agent negligence. It is a story about being able to control and mandate only so much.

An agent who has a listing to launch, a buyer to assist, or a property description to write has a job to do today. If the brokerage has not provided a company sanctioned AI tool, the agent will reach for whatever is free, fast, and available. Most do. A free consumer subscription from any of the major model providers is a thirty-second setup and a meaningful productivity lift.

The exposure is real. Contract information pasted into a prompt. A privacy setting left on its default. Model training opt-outs ignored or misunderstood. Prompt histories stored on personal accounts the brokerage cannot audit. Third-party integrations the agent never read the terms on.

None of this requires bad intent. It requires only that the agent needs to get work done and the brokerage has not told them where to do it safely.

Pressure point three: policy as the forcing function, paired with a sanctioned tool

A written AI policy is the most leveraged artifact in the stack, for one important reason. It is enforceable. Technical controls stop at the perimeter. Training only goes so far even with the most well intentioned plans. A policy, signed or acknowledged, is the document that defines expectation and accountability across the entire agent population.

But a policy that prohibits without providing is a policy that gets ignored. This is where brokerages can easily fail. They publish a document that says do not use consumer AI with sensitive information, and they stop there. The agent, still facing the listing, the buyer and the deadline, does exactly what the policy prohibits, because no alternative exists.

The enforceable policy pairs restriction with a sanctioned path. It says, here is what you cannot use. Here is what you can use. Here is how to use it. Here is what we have done at the admin level to keep your work and your client’s information inside an environment we have vetted.

This is where the stack becomes coherent. Policy defines the rule. The sanctioned enterprise tool, deployed with SSO, admin controls, and reviewed data handling terms, makes the rule followable. Training and reinforcement closes the loop.

Where this conversation is actually happening

Brokerage networking groups and real estate brands gather regularly to share best practices. When the conversation turns to AI, the governance questions are the same ones every serious firm is working through, regardless of size or geography. The sophistication of the room does not change the difficulty of the problem. It only sharpens the focus on which parts of the stack are still open.

The framework above reflects both operating experience running a multi-state brokerage at scale and the discipline of the University of Michigan’s Chief Data and AI Officer program, where governance is treated as foundational rather than optional. In our own environment, the sanctioned path is an enterprise workspace AI deployment, integrated with SSO, administered centrally, and paired with a written policy and agent training. The specifics of the model matter less than the pattern. The pattern is what travels.

The liability window has been open for years

One framing correction for any operator is still treating this as a future problem ChatGPT crossed into mainstream awareness in November 2022. For more than three years, most agents across brokerage firms have been using consumer AI tools on personal devices, often with sensitive data, almost always without governance.

Brokerages still treating governance as a planning exercise are accumulating exposure by the day. The firms building the stack now are not getting ahead of a future curve… they are closing a gap that has been compounding for years.

The diagnostic

Every brokerage leader should be able to answer the following with a yes.

1. Do we have a written AI policy that defines acceptable use, prohibited use, and data handling expectations?

2. Have we deployed a sanctioned enterprise AI tool with SSO, data loss prevention, and admin controls?

3. Have we reviewed the data handling and model training terms of that tool? Does it align with the brokerage’s confidentiality and compliance needs?

4. Do our agents know which tools are sanctioned, which are prohibited, and where the line is?

5. Have we provided AI training resources to agents and employees?

6. Do we have any visibility into what consumer AI tools our agents are currently using with client data?

If any answer is no, the governance stack is incomplete, and the exposure persists.

Enterprise AI licensing is where this conversation starts. It is not where it ends.

I was at Inman Connect in NYC in early February. What I noticed shouldn’t be surprising to anyone: every conversation had AI in it. Not because every company had built something meaningful with AI, but because the letters A and I have become mandatory vocabulary. CRM vendors are now “AI-powered CRMs.” Transaction platforms are “AI-driven.” Marketing tools are “AI-enhanced.”

Brad Inman opened the conference by calling this the “5th industry revolution.” He’s right. But revolutions don’t fail because the technology isn’t ready. They fail because the architecture underneath it can’t support what the technology needs to do.

The Architecture Problem

There’s a number that stuck with me from a conversation with another operator at the conference: 30%. That’s the estimated time agents and staff waste jumping between disconnected systems - an “integration tax” as it was referred to. Not because systems are bad individually, but because nobody built them to work together.

Every vendor wants to fix this. The pitch is always some version of “we’ll be your single pane of glass.” I’ve seen this play out before. Having the idea is one thing. Being able to execute and deliver is a whole other story.

Here’s the deeper issue: many of these “integrated” platforms were built through years of acquisitions and bolt-on partnerships held together by middleware and API duct tape. On the surface, they look unified. Underneath, the data is siloed in different codebases. Rechat CEO Shayan Hamidi called this the “Frankenstein Problem.” These platforms weren’t designed as one system. They were assembled into one.

That distinction matters now more than it ever has. When humans were the ones clicking between screens, integrated-but-fragmented was tolerable. We adapted. We built workarounds. But AI doesn’t build workarounds. AI agents need context to operate effectively. If the software has to jump through hoops just to talk to itself or reveal data points, the AI will always be slow, buggy, and blind to the context it needs most.

When your information is scattered across ten different applications, AI without unified context is just working with partial vision. You can’t automate what you can’t see. Adding AI to that fragmented environment doesn’t create efficiency. It creates faster chaos.

The Adoption Problem Nobody Wants to Discuss

Joe Skousen and his team at Inside Real Estate hosted a breakfast session during the conference. When AI is deployed well within an integrated system, his platform showed roughly 8x more conversations started, 3x more saved searches, and 3.4x more listings favorited. Those are real behavior changes.

But the operative phrase is “deployed well.” And “deployed well” increasingly means deployed within a system where the AI has full context, not just access to one slice of the workflow. The difference between a CRM that tells you a lead went cold and an AI agent that re-engages them automatically.

Most technology in most brokerages is not deployed this way. It’s purchased, announced, lightly trained, and then ignored by 70% of the organization.

I’ve been demoing real estate technology for two decades and the pace of innovation is increasing exponentially. We recently deployed AI-assisted productivity tools across our entire 1,200-person firm. Even with full buy-in and a clean rollout, adoption is a daily discipline, not a launch event. The single most important KPI for any technology decision isn’t the feature set, the demo, or the price. It’s this: how many agents will actually use it?

That’s not a technology question. It’s an operations question.

A friend and leader of another Sotheby’s International Realty firm, mentioned during a panel he participated in that he is deliberately pausing technology investment in 2026 to focus on market growth and agent business building since there are some many things evolving so rapidly right now. We’re at an inflection point as an industry. Consolidation is accelerating, technology is becoming more democratized and accessible almost weekly. A deliberate strategy and vision is more important than even the single best tool.

When you put brokerage operators and proptech founders at the same table without a sales agenda, the conversation shifts completely. Founders start asking what actually breaks in deployment. Operators start explaining the 1099 reality: you can’t mandate tools, you can’t force training attendance, and your best producers are the least likely to change their workflow for anyone.

You can have policies. But you can’t force adoption. You can’t force someone to change a workflow built over years or even decades of habits. The best technology for an agent is the technology they choose to use daily, not because they’re forced to, but because they see the value and leverage it gives them.

What I’d Ask Instead

After days of pitches, panels, and private conversations, here’s the filter I’m applying to every vendor conversation moving forward. Five questions:

Is it integrated, or is it native? There’s a difference. Integrated means bolted together. Native means built together. If the platform was assembled through acquisitions and partnerships, ask how the data actually flows between modules. The answer will tell you whether AI can operate across the full workflow, or whether true context flows between different systems and modules, unlocking the seamless simplicity agents and brokers are actually looking for.

What context does the AI actually have? If a platform’s AI can see CRM data but not transaction history, or marketing performance but not lead engagement, it’s operating with partial vision. The whole point of AI is pattern recognition across the full picture. Anything less is a search bar with better branding.

What’s the adoption path for 1,000+ agents who didn’t ask for this? If the answer involves “training sessions” and “rollout communications,” they haven’t thought hard enough about change management in a 1099 workforce. The best tools don’t require adoption campaigns. They become indispensable because they deliver visible results immediately.

What happens to my data if I leave? Data portability was a recurring theme at Inman. If a vendor can’t answer this clearly, they’re building a moat around your information, not around their product.

Can you show me results from a brokerage my size, in production, for more than six months? Not a pilot. Not a case study from a 50-person team. Real results at real scale with real agents.

The Bigger Picture

The best technology decision I’ve made in 20 years wasn’t picking the right vendor. It was building the internal capability to evaluate vendors honestly, deploy them pragmatically, and walk away when the promises didn’t match the outcomes.

But we’re at an inflection point. The question is no longer “does your platform integrate with everything else in the stack?” The question is “was your platform built so that integration isn’t necessary?” That’s the architectural shift happening underneath all the AI hype. And the brokerages that recognize it early will have a structural advantage that compounds over time.

Nobody is coming to save your tech stack. But for the first time, the right architecture might actually make it possible to build one that doesn’t need saving.

Most people in real estate are collecting prompts.

“Use this one for listing descriptions.” “Try this one for buyer emails.” “Here’s my mega-thread of 50 prompts for agents.”

I get it. I started there too. And prompts work. If you’re an agent, a manager, or a marketer who went from staring at a blank screen to generating a solid first draft in 30 seconds, that’s real progress.

But I want to show you what comes next. Because the distance between “using AI with good prompts” and “building AI into how you work” is where things get genuinely interesting. And you don’t need to be technical to cross that line.

I’m not a developer. I can’t write code from scratch. But I can work alongside AI to build real things. Last Sunday, I spent the afternoon setting up a local knowledge base using a vector database and Claude. The Sunday before that, I built a custom workflow that triages my inbox by urgency and stakeholder priority before I’ve had my coffee.

I’m also not an AI expert. I’m an operator. What I’ve found is that there’s a whole space between “I use ChatGPT sometimes” and “I’m a developer” that almost nobody in real estate is talking about. A space where you build real systems, real workflows, real infrastructure, using plain English and AI as your co-builder.

That’s what this post is about.

The Prompt Ceiling

Here’s the problem with prompt libraries: every time you use one, you’re still doing the work.

You find the right prompt. You paste it into a chat window. You feed it context about this specific property, this specific client, this specific situation. You review the output. You tweak. You paste the result somewhere else.

That’s not automation. That’s a more organized version of doing everything yourself. You’re still the middleman between the AI and your work product.

I hit this ceiling a while back myself. I had dozens of saved prompts. They were good. But I was spending almost as much time managing the prompts as I was saving by using them.

Then I found something that changed the equation entirely.

What a “Skill” Actually Is

Claude, the AI I use daily (and wrote about in my last post), has a feature called Skills.

The concept is simple: A prompt is something you type every time. A skill is something your AI already knows how to do.

You write a plain-English instruction file, a single markdown document called SKILL.md, that describes a workflow. What triggers it. What steps to follow. What rules to respect. You upload it once. From that point forward, Claude follows those instructions automatically whenever the right context shows up.

No code. No JSON. No API keys. Plain English in a text file.

Here’s one distinction worth understanding early: a Project is where your AI stores what it knows about you. A Skill is where it stores what it knows how to do. Projects give your AI context. Skills give it capabilities. Both matter. But if you’re just getting started, Projects come first.

A skill has four parts:

• Description: What this workflow does.

• Trigger Phrases: When it should activate.

• Procedure: The steps, in order.

• Rules: Guardrails so it doesn’t do something you’ll regret.

That’s it. If you can write a checklist for a new hire, you can write a skill.

But Here’s the Thing Most People Miss

Before you ever build a custom skill, Claude already has powerful capabilities built in that most users never discover.

This is where I want every agent, manager, and marketer reading this to pay attention, because what I’m about to describe is available to you right now, today. No setup. No uploads. No technical anything.

Claude can create actual files.

Not text in a chat window that you copy and paste. Actual, downloadable, ready-to-use documents.

Tell Claude: “Create a listing presentation for 47 Maple Drive, a 4-bedroom colonial in Westport listed at $2.85M. Include a market overview, pricing strategy, and 90-day marketing timeline.”

You get a PowerPoint file. A real deck that opens in PowerPoint or Google Slides. Formatted slides. Speaker notes. Ready to customize and present.

Tell Claude: “Here’s my last 6 months of closed sales data. Build me a client-ready market report with median price trends and days-on-market by neighborhood.”

You get an Excel spreadsheet with working formulas. Or a formatted Word document. Or both.

One honest caveat here: Claude can build the deliverable, but it can’t pull live MLS data on its own. That’s the integration layer the industry is still building. I’ve been exploring a platform called Repliers that connects MLS data directly to Claude through an MCP server, essentially a bridge that lets your AI talk directly to an external data source without you copying and pasting anything in between. It’s early, but I’ve gotten it running. When this layer matures, the gap between “AI can help me write” and “AI can build my entire listing launch package from live MLS data” closes fast.

Tell Claude: “Write a just-listed email for my sphere, a property description for MLS, and a social media caption. The property is a waterfront contemporary in Rowayton with 180-degree harbor views.”

You get a Word document with all three, formatted and ready to use.

This isn’t future technology. This is what the tool can do for you today.

What You Should Build First (And It’s Not a Skill)

If you take one thing from this article, let it be this:

Build a Project.

I use Claude, but if you’re a ChatGPT user, set up a project there. The concept is the same. Just get started. The tool matters less than the habit.

A Project is a workspace where you upload documents that give the AI permanent context. Think of it as giving your AI a filing cabinet about your business.

Here’s what to put in it:

• Your last 15-20 listing descriptions and marketing emails

• Your bio and personal brand positioning

• Your active listings with property details and talking points

• Market data for your farm area

• Any templates you reuse: buyer guides, listing presentations, CMA frameworks

• Examples of emails or messages that sound like you

Now every conversation inside that Project starts with Claude already knowing who you are, how you write, what you’re working on, and what your market looks like.

You don’t need to engineer anything. You just talk.

“Write the description for the new listing on Harbor Road. You know my style.”

And it does. Because it’s read 20 examples of your writing and understands your voice, your level of formality, the way you describe architectural details, your tendency to lead with lifestyle over specs (or the reverse).

“Draft a follow-up email to the Millers after Saturday’s showing. They loved the primary suite but were concerned about the commute. Be encouraging but honest.”

Claude writes in your voice, references the specific property and the specific concern, and produces something you’d actually send. Maybe with a tweak or two. But the heavy lifting is done.

“I have a listing appointment Thursday. The sellers interviewed two other agents that I compete with daily. Help me prep three talking points that differentiate my approach.”

Now Claude is pulling from your brand positioning, your track record, your market knowledge, all uploaded in the Project, to help you prepare for a competitive pitch.

This takes about 20 minutes to set up. The return is immediate and it compounds over time as you add more materials.

The Builder Category

There’s a narrative in this industry that you’re either “technical” or you’re not. You either code or you don’t. You’re an AI person or you’re still figuring out what an LLM is.

I think that’s wrong, and I think it’s holding people back.

I spent last Sunday building a vector database on my laptop. I didn’t write the code. I described what I wanted, Claude wrote the scripts, and we troubleshot together when things broke. By the end of the afternoon, I had a working local knowledge base: a perfect memory of every conversation I’ve had across Claude, ChatGPT, and Perplexity, searchable and queryable from inside my AI workflows. No more losing context between tools or starting over every conversation.

I couldn’t have done that alone. But I also couldn’t have done it if I didn’t understand what I was trying to build and why. The AI handles the syntax. I handle the architecture and the intent.

That’s the new category. Not “coder.” Not “user.” Builder.

The more I study the foundational technology, the more convinced I am that this category is where most ambitious professionals will land. You don’t need to understand backpropagation to build a skill that saves your team 10 hours a week. But understanding how these models process context, retrieve information, and generate output makes you a dramatically better builder.

What This Means for Residential Real Estate

The brokerages that figure this out first aren’t just going to save time. They’re going to operate at a level of consistency and speed that becomes a value-add for their agents.

When a prospective agent sees that your platform can produce a custom listing presentation in minutes, that your onboarding is seamless, that your market reports generate themselves, that your AI tools actually know the agent’s voice and market, that’s a value proposition they can feel. Not another CRM demo. Not another training portal. Something that tangibly makes their business better from day one.

And agents should be asking about this. Not “do you have AI tools?” Everyone will say yes. The better question: “What have you built?”

The answer tells you everything about how seriously a firm’s leadership is thinking about what comes next.

Where I’m Going With This

I don’t have this figured out. Nobody does. There’s a saying I’ve carried with me for years: you’re living life in beta. Nothing is ever fully complete. It may be shipped to production, but an update is right around the corner. That’s all of us right now with AI. The tools change monthly. The capabilities expand weekly. The best you can do is stay in motion, build what you can with what’s available today, and be ready to adapt when the next update drops.

That’s partly why BrokerageOS exists.

But I’ve seen enough to know that the gap between “uses AI” and “builds with AI” is smaller than most people think. It starts with a Project that knows your business. It scales with Skills that encode your best practices. And it compounds as you invest the time to understand what these tools can actually do.

Someone needs to explore this openly. To share what works, admit what doesn’t, and pull others forward.

That’s the job I’m signing up for. I hope you’ll come along.

Most people I talk to in real estate default to ChatGPT. It makes sense. OpenAI spent billions on brand awareness. ChatGPT was first to market. It’s literally the Kleenex of the LLM world.

But I’ve been using Claude for years now, and I can tell you with confidence: these are not the same tool. The differences aren’t cosmetic. They’re structural. No matter who you are, understanding those differences matters.

ChatGPT is great at a lot of things. But Claude is where the real work gets done.

Why This Matters Right Now

This week, Anthropic released Claude Opus 4.6, its most advanced model to date. A 1-million-token context window. Agent teams that split complex tasks across multiple AI workers in parallel. The highest scores ever recorded on benchmarks measuring real-world professional work in finance, legal, and other domains.

Days earlier, Anthropic launched a legal plugin for Claude Cowork that automates document review, risk flagging, and compliance tracking. The market’s reaction was swift. Thomson Reuters dropped 16%. LegalZoom fell 20%. The London Stock Exchange Group plunged 13%. Nearly $300 billion in market value was wiped from software stocks in a single day. Traders at Jefferies called it the “SaaSpocalypse.”

The selloff wasn’t about one plugin. It was about what that plugin represented: AI moving from the chatbot layer into the application layer, directly into the workflows where SaaS companies have built their pricing power. The tools are no longer toys. They’re reshaping how professional work gets done.

Credit Where It’s Due

ChatGPT earns its reputation. Real-time web research is fast and fluid. Image and video generation through DALL-E and Sora are strong creative tools. Voice chat with multiple personas is polished.

If your AI use is primarily research, creative brainstorming, and quick technical work, ChatGPT is a solid choice. But that’s not what serious professionals need most.

Start With the Right Foundation

Not every AI tool belongs in every role.

My firm is a Google Workspace client, so we deployed Gemini for Workspace across our entire population. Gemini lives inside the tools our people already use: Gmail, Docs, Sheets, Slides, and Drive. It helps them draft emails, summarize threads, generate listing content, and analyze market data without leaving their workflow.

Most critically, Gemini operates within our secure enterprise environment. Agent communications, client information, and brokerage data stay protected inside our Google ecosystem. For a brokerage with over 1,200+ users across three states handling confidential transactions daily, that security layer isn’t optional.

Gemini for Workspace gives our agents and employees a competitive advantage in their daily production. Claude rivals that for corporate use, and in many ways goes further.

Where Claude Separates Itself

When you need depth, context, and a tool that can hold the weight of real business problems without losing the thread, Claude and more specifically Claude for Teams is in a different class.

Deep document analysis. I regularly upload leases, compliance datasets, vendor proposals, and reports and work through them in ways that would have taken hundreds of hours manually. We’ve run compliance analysis across thousands of agreements, identifying patterns and gaps that would have been buried in spreadsheets for weeks. Both tools offer large context windows now, but Claude maintains coherence and precision across long, complex documents in ways that consistently outperform. A 50-page contract, a dataset with thousands of rows. The quality holds up start to finish.

Writing quality. I’ll say it plainly and many agree: Claude writes better. Not flashier. Better. More structured, more precise, more aligned with the way you actually communicate. Independent reviews consistently describe Claude’s tone as warmer, more contextual, and less robotic. When I’m drafting communications, internal memos, or other messages, Claude produces work that needs less editing.

Projects with real depth. Both platforms offer projects, but the customization is where Claude pulls ahead. You can upload extensive reference documents, build custom instruction sets, and create persistent workspaces that shape behavior at a granular level. We’ve built shared projects for compliance analysis, vendor evaluation, and agent content creation. Each carries its own context and rules. Claude already knows your standards, your voice, and your operating parameters before you type a word.

One of the most powerful applications: agent-specific projects that function as digital twin copywriters. Upload collections of an agent’s prior writing, newsletters, emails, and marketing materials, and you can generate content that matches their voice precisely. Support staff produces on-brand content at speed. For top producers who would rather spend time with clients than writing marketing copy, this is a meaningful unlock.

Memory that actually retrieves. ChatGPT has memory, but in practice it captures high-level preferences and themes. It knows you prefer bullet points or that you work in real estate. What it can’t do is search back through a conversation from three weeks ago and pull the exact analysis you ran. Claude can. I reference a prior vendor comparison, a compliance review, a draft I worked through weeks ago, and Claude searches across past conversations by keyword or time range and returns the actual content. That’s the difference between a tool that remembers your preferences and a tool that remembers your work.

Extended thinking. Claude reasons through complex problems step by step before responding. With Opus 4.6, this evolved into adaptive thinking, where the model dynamically decides how deeply to reason based on the task. I’ve used this for financial modeling, strategic planning, and multi-variable decision analysis. The depth is materially different from a standard response.

Cowork. This changed how I think about what AI can do for a professional. Launched in January, Cowork is fundamentally different from chatting with an AI. You point Claude at a folder on your local computer, describe what you need, and step away. It plans, executes, and delivers finished work: organized files, formatted spreadsheets, polished presentations, synthesized research.

Chat tells you how to do something. Cowork does it. And because it runs locally on your machine with access only to the folders you choose, your data stays where it should.

Cowork is superior to anything I can accomplish in standard chat or with Gemini for Workspace right now. It’s early. It’s a research preview. But the trajectory is clear. This is where professional AI is headed. Not prompting and pasting. Delegating and reviewing. That applies whether you’re organizing files, building a market report, or prepping a presentation.

Claude for Teams

We deployed Claude for Teams across staff in finance, operations, and communications. Shared projects, privacy protections, and collaborative features made the decision straightforward. The tool doesn’t just help individuals work faster. It raises the floor for the entire team.

We also use Claude for Excel for rich data analysis on complex spreadsheets. This isn’t leadership experimenting with a new toy. It’s team members integrating AI into their daily workflows because it makes their work materially better.

The Governance Factor

This matters more than most people realize, especially in a regulated industry like real estate. While both OpenAI and Anthropic now use consumer data to train their models by default, the distinction lies in the Team and Enterprise tiers.

We tested out Claude for Teams because it provides an immediate, legally-backed wall between our proprietary data and the public model. When you’re uploading sensitive agreements or financial modeling, “Opt-out” shouldn’t be a setting you hope your team remembers to toggle, it should be a structural guarantee of the platform. For a brokerage, the choice isn’t just about the AI’s “brain”; it’s about the “vault” it sits in.

Level Up

The professionals who gain real leverage from AI won’t be the ones who adopted it first. They’ll be the ones who adopted it deliberately.

What happened in the markets this week should be a wake-up call. AI is no longer a productivity experiment. It’s becoming the operational layer for professional work. The tools are moving fast, and the people who understand which tool to use, when, and why will have a meaningful advantage over those still defaulting to whatever they heard about first.

I’ve watched Claude evolve from a lesser known alternative into the most capable tool in my daily workflow. If you’re already using ChatGPT and want to go deeper, Claude is the logical next step. Not because ChatGPT is bad. Because once you experience the depth of what Claude offers, you won’t want to go back.

The strategy isn’t picking the AI everyone knows. It’s picking the right tool for the job you actually need done.

The transaction created the largest residential brokerage in U.S. history: 340,000+ agents, six legacy brands including Coldwell Banker, Century 21, and Sotheby’s International Realty, plus integrated title, escrow, and relocation services under one roof.

This follows Rocket’s acquisition of Redfin last year, new ways of doing business and working with customers taking shape, and existing home sales sitting at their lowest level since 1995.

The industry is expanding, contracting, and determining its destiny in real time. Nothing is sacred. Everything is changing and that’s why I got the compulsion to show up here and start writing.

Who Is This Guy?

I grew up around real estate. An aunt who still practices today and is affiliated with our firm. A family of contractors. But I didn’t grow up wanting to run a residential real estate brokerage. I had other aspirations. Things change, like they do for many, and real estate had a way of finding me. Right place, right time, and the rest is history.

I don’t code. I’m not a traditional tech person. But I’ve been tinkering since an early age, seeing how technology could be deployed to improve systems and processes, always looking for what’s next. How to improve an agent’s day to day by simplifying workflow so they can focus on what matters most: building relationships.

I don’t fit neatly into one box. I understand tech, cross disciplines, run operations, and have gone from an admin in a real estate office to Co-President and COO of William Pitt Sotheby’s International Realty and Julia B. Fee Sotheby’s International Realty, the third-largest Sotheby’s International Realty firm in the world and consistently ranked as one of the top 30 brokerages in the US by industry publications.

Thirty offices across Connecticut, New York, and Massachusetts. 1,100+ agents. $6+ billion in annual volume. We’ve been independently owned and operated for 76 years and remain that way today, no matter what is changing around us.

More than two decades in the industry later, it’s been an incredible journey. I’ve spent that time learning what it actually takes to run a brokerage: navigating complex transactions, supporting agents who wake up unemployed every day until they’re not, helping entrepreneurs realize their potential by unlocking technology that works for them rather than against them.

Brokerage OS is where I’ll share what I’ve learned in my career, what I’m learning right now, and what I think is worthy of putting to paper (proverbially speaking of course). These aren’t conference talking points. It’s the practical work of being a brokerage operator when the ground is moving.

What the Consolidation Signals

The Compass - Anywhere deal matters less for its size than for what it reveals about where leverage is shifting in our industry.

Scale used to be about market coverage and recruiting capacity. Now it’s increasingly about technology platforms, vertical integration, and control over listing distribution. Compass isn’t just building a bigger brokerage. They’re building infrastructure that could challenge the institutions themselves: the MLS, the portals, the traditional rules of how inventory flows.

But consolidation isn’t the only force reshaping this industry. AI is at the doorstep ready to shake up everything and you can already feel its presence.

The way systems run in this business are changing and going to keep changing. Compliance. Deal flow. Lead nurturing. A level of automation and true value creation that everyone from a small one-office brokerage to the biggest heavyweights will have access to. For the first time, capability won’t be gated by scale or capital. It will be gated by who understands how to implement it.

The operating questions this raises apply across the industry. Which variables are changing? How fast? Agent economics. Listing visibility. Technology investment. Referral network dynamics. None of these have settled into a new equilibrium yet. The operators paying attention to their own data will see the shifts before industry reports tell them what happened.

What I Believe

Scale alone doesn’t win. Operating discipline and adaptability do.

AI is accelerating this reality. Not the hype cycle version that dominates conferences, but the practical application: workflow automation, knowledge retrieval, agent enablement tools that reduce friction instead of adding it. I’m currently enrolled in the Chief Digital and AI Officer program at the University of Michigan because I believe the next generation of brokerage leadership needs to understand these systems at a foundational level, not delegate them to a vendor or an IT department.

The brokerages that thrive over the next five years won’t necessarily be the largest or the best-funded. They’ll be the ones that understand their own economics with precision. The ones that build systems agents actually adopt. The ones willing to redesign workflows, rethink vendor relationships, and question assumptions that made sense five years ago but don’t anymore.

This work requires flexibility. Knowing when to adapt. Seeing opportunities when they present themselves rather than waiting for certainty that never comes.

What’s Coming

This newsletter will cover operations, technology, strategy, and leadership from an operator’s perspective. Commentary grounded in what I’m seeing, not speculation from the sidelines. What’s actually working, what’s broken, and what decisions I’m facing in real time.

Let’s go.