Hey there, folks. There's a word showing up everywhere in real estate tech right now, and it's worth paying attention to: infrastructure. Not tools, not assistants — infrastructure. A major California brokerage just rebuilt its entire operation on top of an AI operating layer, where agents interact with the system by calling a phone number and talking to it like a colleague. Lone Wolf is embedding AI directly into the platforms brokers already use, aimed squarely at the industry's most expensive problem: agent churn. But alongside that momentum comes a sobering data point from the title world — when researchers tested AI title search on 200 real residential files, it missed at least one meaningful title issue in 40.8% of them. The floor is being rebuilt. Just know where the soft spots are. Let's dig in.
1. A $6 Billion Brokerage Just Rebuilt Its Entire Tech Stack on an AI Operating Layer
FirstTeam Real Estate — a California independent brokerage that did $6.12 billion in volume across 5,978 transactions in 2025, ranking 34th nationally by volume — is standardizing its entire technology stack on Purlin, an AI operating layer that already counts more than 40,000 agents, teams, loan officers and brokerages as users across North America. The rollout covers three products: PurlinOS (the operational core), Purlin Close (which automates contract execution and related workflows), and Purlin Offer & Negotiate (live, transparent bidding).
The detail that stands out is the interface. FirstTeam agents can work the system by voice, text, email or chatbot — including calling a phone number and talking to an AI assistant the same way they'd call a colleague. Purlin's chief revenue officer, Tim Quirk, said that matters because picking up a phone is something every agent already knows how to do, which removes the adoption friction that kills most brokerage tech rollouts. FirstTeam's VP of marketing and strategic initiatives, Lauren Henss, told Inman the brokerage chose Purlin over existing all-in-one platforms specifically because the company was willing to build custom functionality alongside them rather than hand over a fixed menu of features. In a separate HousingWire interview, Henss and Quirk were notably measured about limits — emphasizing that successful AI integration runs on education, agent ambassadors, utilization metrics and leadership buy-in, and that AI should support but not replace agent expertise, especially in negotiations and local market knowledge.
Why It Matters: This is what the "agentic" shift looks like when a real brokerage commits to it fully rather than bolting on another tool. Two things are worth your attention. First, the voice interface is a genuinely smart adoption play — the single biggest reason brokerage tech fails isn't capability, it's that agents never learn it. A system you can simply call and talk to sidesteps the training problem entirely. If you're evaluating AI tools, weight ease of adoption as heavily as feature lists; the most powerful platform your agents never open is worth nothing. Second, note the organizational detail buried in the HousingWire piece: education, agent ambassadors, utilization metrics, leadership buy-in. That's the unglamorous checklist that determines whether an AI rollout works, and it's the part most brokerages skip. If you're a broker-owner weighing a big AI investment, the technology decision is the easy half.
2. Lone Wolf Aims AI at Real Estate's Most Expensive Problem: Agent Churn
Lone Wolf Technologies is making a different bet — that the next phase of AI in real estate isn't a standalone product but software that works quietly inside the platforms brokers already use. The company unveiled Talent, an AI-powered module for its BrokerMetrics analytics platform, alongside Aspen, an AI assistant designed to work across Lone Wolf's entire software ecosystem.
Talent points AI at recruiting and retention. The assistant is built to identify recruiting prospects, flag flight-risk signals among current agents, and answer broker questions in plain English rather than requiring someone to build a report. New CEO Matt Fischer framed the problem candidly: brokers have long had the BrokerMetrics data, but turning that information into action has been the hard part. Notably, Lone Wolf's own 2026 survey found brokerages rated retaining agents significantly more valuable than recruiting new ones — a meaningful shift in priority. In a related move signaling where the industry's data plumbing is heading, property data platform Repliers formed a partnership with Unlock MLS to give Austin-area professionals API access to Central Texas market intelligence for AI-assisted development. As Repliers CEO Rhett Damon put it, AI products are only as good as the data they can reach.
Why It Matters: The retention angle deserves more attention than it's getting. NAR data has long shown the majority of agents leave the business within a few years, and every departure costs a brokerage real money in onboarding, training, and lost production. If AI can reliably flag a flight-risk agent before they walk — based on activity patterns, production trends, and engagement signals — that's a genuine dollars-and-cents win for broker-owners. But there's a flip side worth naming honestly: if your brokerage deploys tools like this, your own activity data is being analyzed to predict whether you're thinking about leaving. That's not sinister, but it is new, and agents should understand it's happening. The broader pattern here matches what we've tracked all year — the industry is standardizing on AI that's embedded in existing workflows rather than sold as another separate subscription. That's the right direction, because the standalone-tool era produced a lot of software nobody opened.
3. AI Title Search Missed Real Issues in 40.8% of Files — and Couldn't Search 8% at All
Now the counterweight, and it's a significant one. DataTrace Information Services released a study — "AI Title Search Tested in the Real World: What Accuracy, Risk, and Readiness Really Look Like" — testing whether AI relying on public records alone can produce the accuracy required for insurable title decisions. Reviewing 200 randomly selected residential title files, public-record-only AI search missed at least one meaningful title matter in 40.8% of searchable files when compared against searches supported by title plant data.
The failure pattern matters as much as the headline number. AI couldn't complete a search at all on 16 of the 200 files (8%), because it lacked the title plant data or comparable normalized datasets needed. And the misses weren't evenly distributed — the most significant gaps occurred in high-risk categories, with involuntary liens showing an issue fail rate above 36%. DataTrace's chief data officer, Annette Cotton, framed the conclusion carefully: AI performs best when operating on structured, validated title data rather than fragmented county-level public records, and the speed AI offers creates value only when paired with completeness and accuracy. A related DataTrace paper made the scale implication vivid — a 1% variance in data accuracy applied to 5 million annual transactions could produce up to 50,000 instances of inaccurate title. Worth noting the source's interest: DataTrace sells title plant data, so the study validates its own product. But the methodology is concrete and the file-level findings are specific enough to take seriously.
Why It Matters: Pair this with the Columbia mortgage study we covered in August — where leading models scored between 51% and 80% on basic loan-file verification — and a consistent, important pattern emerges. AI is impressively fast at retrieving documents and dangerously unreliable at concluding things from them, especially when the underlying data is fragmented. Public records were built to provide notice, not to verify a title outcome. For agents, the practical read is this: as AI-powered title and closing services proliferate and compete on speed and price, ask what data they're actually running on. "AI-powered title search" built on public records alone is a meaningfully different product from one built on validated title plant data, and your client is the one carrying the risk if something gets missed. A missed involuntary lien isn't a minor inconvenience — it's a cloud on title that surfaces at the worst possible moment. Speed is a feature. Insurability is the product.
That's the wrap, folks. The theme is consistent across all three: AI is moving from a tool you pick up to the floor you stand on — brokerage operating layers, embedded recruiting intelligence, automated title search. That's mostly good. But the title study is the reminder worth carrying: the quality of AI's output is capped entirely by the quality of its data, and in the places where real money and real risk live, that gap still has teeth. Build on the new floor. Just check where it's solid. See y'all next time.
Disclaimer: AiRE Update is an independently produced newsletter that curates and summarizes publicly available news. I don't write the original articles featured here — I summarize them in my own words and add commentary on why they matter. All original reporting, content, and intellectual property remain the property of their respective authors and publications, including Inman, HousingWire, Real Estate News, Commercial Observer, and DataTrace Information Services. Each story links back to its original source, and I encourage you to read the full articles there. The summaries and opinions in AiRE Update are my own and are provided for informational purposes only; nothing here should be taken as legal, financial, or professional advice.