Inside Real Estate wants an artificial intelligence assistant checking your transaction files before they ever reach a broker’s desk. The bigger question circulating among industry watchers isn’t whether that assistant works, but who ends up owning what it learns.
The company recently rolled out ComplianceAI, a new tool built with real estate compliance automation firm SetWave that scans transaction packets inside BoldTrail BackOffice for missing forms, absent signatures and inconsistent information before a deal closes.
Inside Real Estate paired the launch with early access to Model Context Protocol (MCP) support, an open standard that lets AI assistants like Claude and ChatGPT interact directly with a brokerage’s transaction data rather than simply answering questions about it.
Joe Skousen
“Brokerages shouldn’t have to choose between a powerful platform and compliance they can trust,” said Inside Real Estate CEO Joe Skousen in a statement announcing the launch. “ComplianceAI brings that compliance directly into the workflow our customers already use, catching issues before closing instead of after. And MCP is the next step: a brokerage where an agent can open a transaction, prep the paperwork, and get it signed just by asking, all inside the system they already run on.”
According to Victor Lund, managing partner of WAV Group, CEO of RE Technology and CEO of Fluente, the more consequential story running in the background is that, in the industry today, every brokerage now has to decide whether to let its software vendor build the AI system, or build the AI itself.
It is creating a new twist to the age-old proptech and enterprise software debate over “build versus buy.”
The sovereignty question
Lund frames the decision brokers now face as a split between two paths: Build a proprietary AI system trained only on your own data, contained on your own servers, or adopt the AI your existing vendor is already building into its product.
Victor Lund
“As a broker, I’d be nervous, because I have to trust somebody using AI on confidential information to do it right,” Lund told Inman. “I have a high degree of certainty that Inside Real Estate does it right. But I would never trust a startup with that data.”
The distinction matters more than it might sound, Lund argued, because of what he called the training asset. Every time an AI compliance tool flags or misses a problem in a transaction file, that interaction refines the model.
Over enough transactions, the tool gets sharper. But if a vendor owns that model, the broker doesn’t take the intelligence with them if they switch platforms.
“That knowledge belongs to BoldTrail, not the broker,” Lund said. “If you leave, you don’t take it with you, even though you put the work into training it.”
Lund calls the alternative sovereignty: owning the governance layer that defines what the AI can and can’t do, the security layer that keeps transaction data from leaking outside its intended system, and the accumulated intelligence itself.
Compass, he said, has treated that as a strategic priority, pouring massive resources into proprietary AI architecture under the direction of co-founder and chief strategist Ori Allon.
Betting on the vendor
Most brokerages won’t build their own, Lund said. That’s not because the technology is difficult, but because it requires in-house technical expertise most shops don’t have.
He pointed to Seven Gables Real Estate, a roughly $2 billion-volume independent brokerage in Southern California, that built its own AI tools in-house, as an exception rather than the rule.
FirstTeam Real Estate, another large indie brokerage in California, also recently decided to build its own all-in-one AI system for its agents in partnership with Purlin, as Inman recently reported.
For everyone else, the calculation looks more like the one facing agents at eXp, Side, Real and Fathom. These brokerages Lund grouped under a “brokerage as a service” model, where the company supplies both the tools and the compliance oversight layer that keeps agents’ transactions in line.
Inside Real Estate is positioning ComplianceAI to play a similar role for the brokerages and franchise networks running on BoldTrail.
Whether that trade-off bothers real estate agents as much as it does brokers is a separate question. “I don’t know how much agents care. Agents just want a clean file,” Lund said.
The file-review cycle he described, where a broker flags five missing items, the agent fixes four and introduces a new problem with the fifth, is the exact friction ComplianceAI and tools like it are designed to compress.
“If AI can take more of that work off the agent’s plate — instead of ‘I need this document,’ it just goes and gets the document — that’s the evolution happening,” Lund said.
Lund continued, “I don’t know how good the BoldTrail solution is yet; no brokers have given me feedback, since it just launched. I’m sure it’ll start out OK, but given the volume of transactions BoldTrail already manages, it’s going to get smart fast.”
‘Think of it as an AI talking to an AI’
The sovereignty debate extends beyond compliance review into how brokerages stitch their broader tech stacks together.
Inside Real Estate is also building out Streams Studio, an MCP-native platform the company describes as connecting a brokerage’s full technology stack of compliance, transactions, marketing and lead generation rather than just BoldTrail BackOffice.
Lund compared it to an architecture Berkshire Hathaway built independently, pairing an AI interface called Maestro with a connective layer called Event Bus that keeps Rechat, its CRM platform, Reliance, and its commission-management tool AccountTech synced without duplicate data entry.
He said the approach is best-of-breed point solutions, connected by an AI agent that manages workflow across all of them, plus an Event Bus that lets the AI “listen.”
If something happens in a transaction in Rechat, it’s sitting there listening and says, “I need to update the CRM; I need to notify the transaction manager.” The Event Bus does that, so the agent isn’t duplicating entries or sending redundant emails.
The BoldTrail version of that idea, Lund said, is meant to work across point solutions a brokerage already uses, such as MoxiWorks, Follow Up Boss, kvCORE, rather than requiring a full-suite switch.
“Think of it as an AI talking to an AI,” Lund said. “And again, it’s secure, intelligence happens at the AI level, it follows business rules, and it gets better over time. Every failed request trains it.”
The MLS layer is moving in a similar direction, Lund said. He points to MCP servers under development at NorthstarMLS and REcore Solutions that route routine queries through contained, open-source models rather than premium subscriptions to Claude or ChatGPT. This reserves the frontier models, in his view, for work that actually needs them.
“A lot of people using these frontier models are asking them to do the work of a first grader,” Lund said. “These models are smarter than a team of Ph.D.s from every university in the world. You don’t need that to build a CMA. A pretty basic tool can put that together. But the learning part: that’s the real asset.”
What it means for brokers now
None of this changes what a broker has to do this week, such as review files, close deals, and keep agents compliant. What it changes is the leverage brokers hold over their own data and workflows going forward.
A broker who adopts ComplianceAI today is betting that Inside Real Estate’s governance and security practices are sound enough to justify not building the same thing independently. It’s a bet Lund said he’d make, but wouldn’t extend to an unproven vendor.
“It’s not hard to do. You just have to know how, and do it right,” Lund said of building a properly contained AI system. The industry, by his account, is still years away from settling on which approach wins.
