BCREA’s AI Brokerage Guidance: Six Speaking Points, Six Technical Requirements

Most AI guidance in real estate is written for individual agents. The British Columbia Real Estate Association wrote its edition of Brokerage Meeting in a Box for the person who decides what tools a brokerage allows: the managing broker. The format is a set of speaking points and linked resources for a brokerage meeting, and BCREA offers to send one of its Professional Services Support Advisors, Jim McCaughan or Marty Douglas, to present the material online or in person.

The choice of audience is the point. In British Columbia, the managing broker is the licensee the regulator holds responsible for supervision under section 28 of the Real Estate Services Rules. A speaking-point package aimed at that role turns AI from an individual habit into a brokerage policy decision, which is exactly where BCFSA’s own guideline says it belongs. The key takeaway BCREA gives brokers to close the meeting is that AI can improve efficiency and support the work, but the REALTOR® remains responsible for accuracy, confidentiality, transparency, and professional judgement in every client interaction.

The points are short enough to read aloud. We reproduce their substance here because the rest of this note argues that each one is a technical requirement in plain clothes.

Two things stand out. First, the list is almost identical in substance to CREA’s national principles (transparency, accuracy, accountability) and to the California DRE’s bulletin (AI as an assistant under supervision, verification, disclosure, confidentiality). Three bodies in three jurisdictions, none of them coordinating, arrived at the same six ideas. Second, the second point is the most operational sentence any Canadian association has published on AI: treat AI tools as potentially public environments unless confirmed otherwise. That is a data-classification rule, and it has a direct architectural answer.

BCREA’s fourth point says AI use must align with the BC Financial Services Authority’s rules. The document it means is BCFSA’s Artificial Intelligence Guideline , published in February 2024 alongside an information page on what AI is and how licensee duties apply to it. Its opening position is the one every regulator has since echoed: using AI does not exempt a licensee from legal and ethical obligations, and licensees remain fully accountable for their actions.

What makes the BCFSA guideline unusual is its specificity. It cites the exact Rules that AI use engages and works through four areas of concern with examples drawn from practice.

The guideline also raises a data-use concern that most AI policies miss. It notes that vendors offering AI services such as strata-document review could reuse the incidental data those documents contain for insurance, marketing, or competitive analysis, and it asks licensees to consider that before adopting a tool. In LAIRE terms, that is a retention and secondary-use question, and it belongs in the vendor contract, not in a privacy policy nobody reads.

BCREA aimed its resource at managing brokers because BCFSA did first. The guideline’s managing-broker section reads like a brokerage AI policy outline, and it is the reason a brokerage meeting, not an agent newsletter, is the right venue.

• Brokerage policies: establish clear policies and procedures for AI tools and client information, including protocols for data retention, encryption, and secure sharing, and emphasize anonymizing personally identifiable information before anything enters an AI system.
• Risk mitigation: treat AI output as a supplement to licensee expertise, not a replacement; require referrals for independent professional advice outside a licensee’s expertise; and discuss AI use with the brokerage’s insurer to confirm coverage.
• Educate licensees: make sure every licensee understands the risks, the importance of protecting client information, data privacy regulation, confidentiality, and the proper use of AI tools.
• Vendor due diligence: vet third-party AI tools so their data handling aligns with brokerage privacy policies, client agreements, and obligations under RESA; review terms of service and privacy agreements.
• Ongoing supervision: continuously monitor how licensees use AI tools and regularly review AI-generated content for policy alignment and accuracy.

The last item is the hardest and the most important. Continuous monitoring of how licensees use AI, and regular review of what it produced, is impossible if the tools leave no record. A managing broker cannot supervise a consumer chatbot session that happened on an agent’s phone. Supervision at the standard BCFSA describes requires the tools themselves to report what they did, for whom, on which data, and with what approval. That is not a policy feature. It is a product feature.

A licensee cannot personally deliver most of what BCREA and BCFSA ask if the software does not make it possible. Here is the translation from speaking point to system property, which is how we read every regulator’s guidance before it becomes a control.

There is a reinforcing trend from outside the province. The frontier model providers now enforce most of these requirements themselves. Anthropic’s usage policy classifies housing eligibility and legal interpretation as high-risk uses that require review by a qualified professional and disclosure to the consumer, and prohibits presenting AI output as human. OpenAI’s October 2025 update prohibits tailored legal advice without a licensed professional and automated high-stakes housing decisions without human review. Their browser agents refuse to bypass CAPTCHAs or handle credentials and hold the user responsible for sites’ automated-access terms. A brokerage tool that ignores BCREA’s speaking points is now also fighting its own model provider, and it loses that fight one refused task at a time. Compliance has become the interface through which capable AI agrees to work.

Our LAIRE framework (Layering Agentic Intelligence on Real Estate, formerly RAILS) is built on capability without custody: the AI may be granted narrow, identity-bound, auditable capabilities while client, listing, and transaction data stay in the systems that own them. Version 2.0, published today, adds a regulatory-alignment section and a British Columbia jurisdiction profile skeleton. The mapping to BCREA’s six points is one-to-one.

If you are a managing broker in BC: run the meeting. Then write the policy BCFSA describes: approved tools, prohibited inputs, consent language for client data, labelling rules for altered media, review before release, a vendor due-diligence checklist, and a supervision routine. Ask every vendor one question before anything else: can you show me, per licensee, what your tool did, on which data, and who approved it? If the answer is no, you cannot meet BCFSA’s ongoing-supervision expectation with that tool, whatever its other merits.

If you are a licensee: treat BCREA’s second point as a hard rule until your brokerage confirms a tool otherwise. Nothing that identifies a client goes into a tool your brokerage has not approved. Label staged and enhanced media. Tell clients when a chatbot answered for you. Verify every fact in AI-drafted copy against the listing record before it is published under your name and Rule 41.

If you build technology: build to the BCFSA standard everywhere, because it is the most specific in Canada and it matches CREA’s principles and the California DRE’s bulletin almost line for line. Implement approval gates, minimization, provenance, jurisdiction profiles, human escalation, disclosure events, and supervision exports as product features. That is how we built Homies, and it is why a brokerage meeting like BCREA’s reads to us as a product walkthrough rather than a warning. We hope the rest of the industry builds the same way.