Taylor Avakian runs The Group CRE, a multifamily investment sales team under Lyon Stahl Investment Real Estate in Los Angeles. Since 2018 his team has closed close to $500 million in LA multifamily transactions, and the LA Times has called him a commercial real estate visionary. What separates his practice from most brokerage teams is how he has rebuilt the standard parts of the job around AI: the broker opinion of value, the CRM, his own learning process, and his content. This is a summary of what he built and how other brokers can copy it.
The challenge: standing out in the comparison business
Brokerage, as Taylor describes it, is the comparison business. An owner deciding who lists their building often talks to five teams, and absent a real difference, picks on price or on who they like most. “If you look, sound, and smell like everyone else, then you’ll be perceived like everyone else,” he says, quoting an old business coach. The traditional pitch makes that worse. The broker opinion of value has been delivered the same way for 30 or 40 years: a static three-page PDF that lands in an owner’s inbox looking like every other broker’s three-page PDF. The job is also repetitive by nature, heavy on aggregating sales comps, rent comps, and conversation history by hand.
Play one: an interactive BOV the owner can actually use
Taylor’s answer to the static deck is a live, interactive BOV web page, built with Claude Code. Instead of a fixed PDF, the owner opens a website and changes the inputs themselves: rent growth, NOI, the cap rate, and watches the valuation move in real time. He frames it as proof over promise. When he tells an owner he is technology-forward, the interactive BOV is the evidence in the same breath, and it breaks the pattern of identical decks arriving from every competitor. “This was my way of standing out in front of a sea of sameness,” he says. The page is live and in use.
He is candid that the build was a learning curve. Six months ago he could not have said what a CLI was. Now he works across command-line interfaces and GitHub, and describes the shift as bringing a spark back to a job that can otherwise feel monotonous.
Play two: a CRM you can hold a conversation with
Behind the pitch sits the data. Taylor connected his HubSpot CRM to frontier models, Claude and ChatGPT, through MCP connections and APIs, so he can query his own pipeline in plain language. He uploads and categorizes sales data, pulls in public records, and connects his Gmail and calendar, building what he calls a third brain that is hyper-specific to LA multifamily and to his own book.
The point is the questions it can answer. “I want to pull the top 15 motivated people based on the language they’re using in our conversations,” he says. Who has he not followed up with. Which owners have been active historically. For a broker running 100 conversations a week, that recall is the difference between a lead worked and a lead lost. The value sits in the proprietary data: his notes, his comps, his conversation history, none of which a generic model could reach on its own.
Play three: how he actually learned
Taylor’s advice to brokers who do not know where to start is concrete. Follow people sharing real workflows on X, LinkedIn, and in communities. Download the tools and click around. Ask the model itself what a command does, since it will teach you. Then block real time, two and a half hours on a Saturday, and build one thing start to finish. “You have to actually do the thing,” he says. The mistake he warns against is trying to automate everything at once; the move is to pick a single task and see it through. In his framing, the only real limit now is creativity, because so much of brokerage is already process, steps, and loops, which is what AI handles well.
Play four: AI as a voice amplifier, not a content mill
On social media, where Taylor has built a sizable audience and hosts the No Vacancy podcast, he uses AI as a research and brainstorming partner rather than a writer. Before interviews he deep-researches an owner’s historical sales activity, surfacing context that is hard to find on the public web and sharpening the questions he asks. When a post does not feel clear, he works through hooks and lines until one sounds like him. He draws a firm line at AI-generated slop, which he defines as content in the model’s voice instead of yours. “I want to use AI to magnify and improve my voice,” he says. The reach is real: a stranger recognized him by voice alone at his sister’s college graduation.
What other brokers can take from this
The throughline is that Taylor rebuilt the parts of the job he already owns with general-purpose AI, rather than waiting for a CRE-specific platform to do it for him. The replicable plays: turn your highest-stakes deliverable into something interactive, so the deliverable itself differentiates you; connect your CRM to a model so your proprietary data becomes queryable; start with one workflow and give it real time; and use AI to amplify your own voice rather than replace it.
He is honest about the ceiling. There is no formula that wins business every time. What the interactive BOV and the queryable pipeline change is the odds, by making an owner stop and explore his team instead of filing him with the rest. The page is live, the CRM is answering questions, and both were built with tools any broker can download today.
