AI Tuesdays · Week 07 · How an AI Council Actually Works
MULTIFAMILY NEXT · AI TUESDAYS
Week 07 · How an AI Council Actually Works
Self-guided · ~20 min · Phase 1 Finale
Warmup
The Workshop
Power Preview
Resources & Tools
Take Action
Week 07 of 52 · Self-Guided · ~20 min
0 starred0 notes
Warmup · Week 06 Recap
Welcome back. Let us warm up first.
Last week you learned the 5-Second Check5-SECOND CHECKA quick five-second scan you run before pasting anything into an AI tool. You look for five things (the "5 Ps"): People, Places, Payment, Protected class, Private files. If any of the five show up in what you are about to paste, you stop. Redact if it is Tier 2 data. Refuse if it is Tier 3. Multifamily example: a leasing professional is about to paste a resident application into ChatGPT to draft a welcome email. She runs the 5-Second Check first. Names (People) and apartment numbers (Places) are in the paste. She stops. She replaces both with [RESIDENT NAME] and [APARTMENT NUMBER] placeholders, then pastes. AI drafts the template. She fills the real details back in offline. Taught in full in Week 06. Anchor: RETTC Pillar 4 (Protect Privacy & Promote Data Integrity).. Three quick recap questions to bring it forward. Tap any answer to reveal the coaching, no wrong tries counted.
New to AI Tuesdays or missed last week?
Read Week 06 first for the full context. It teaches the 5-Second Check, which this workshop builds on directly.
1What are the five things you check in five seconds before you paste?
People, places, payment, protected class, private files. If any of the five show up in the paste, stop. RedactREDACT / REDACTIONReplacing sensitive information with generic placeholders before you paste anything into an AI tool. The AI still gets the pattern it needs, but never sees identifying, financial, or protected-class information. Multifamily example, before and after. Before (Tier 2 data, not safe to paste): "Draft a renewal letter for Jane Doe in apartment 214B. Current rent is $1,850, we are offering a 5 percent increase." After (redacted, safe): "Draft a renewal letter for [RESIDENT NAME] in [APARTMENT]. Current rent is [CURRENT RENT], we are offering a 5 percent increase." You paste the redacted version. The AI drafts the template. You fill the real details back in yourself, offline. The resident name, apartment number, and rent never leave your desk. Anchor: RETTC Pillar 4 (Protect Privacy & Promote Data Integrity). if Tier 2. Refuse if Tier 3.
2If the check catches a problem, what do you do next?
Redact, refuse, or route to Council. When in doubt, treat it as YellowYELLOW REQUESTAn AI request that needs Council approval before it runs. Yellow means one of two things: the data is Tier 2 (internal-sensitive or resident-adjacent information like names, apartment numbers, ledger notes, or vendor pricing that needs redaction before going into an approved enterprise AI tool), or the use case is Yellow (an AI task allowed only with guardrails, like drafting resident-facing content or anything that could touch a Fair Housing decision). Both meanings route to the same place: the AI Council. Multifamily example: a property manager wants AI to draft renewal letters that include the resident name, apartment number, and current rent. The data is Tier 2 (personal identifiers) and the use case is Yellow (resident-facing output). Both flags say Yellow. Both say: send to the Council for approval with guardrails. Anchors NIST AI RMF Govern function. and send it to the AI CouncilAI COUNCILThe small group at your property, region, or portfolio that approves Tier 2 AI use cases, refuses Tier 3 requests, and reviews new AI tools before anyone uses them. Standard seats: Governance Lead (chair), Operations, Compliance, Technology, and a rotating practitioner voice. Also called "Governance Team." Multifamily example: a leasing professional wants to use AI to draft renewal notices. Instead of guessing whether it is allowed, she submits the use case to the AI Council. The Council approves it with guardrails (no resident names, no ledger data, human sign-off on every draft) and logs the decision with a 90-day review date. That is the Council doing its everyday job. Anchor: RETTC Pillar 1 (Establish an Organizational Philosophy on AI Use and Oversight).. That is the safest reflex.
3Whose job is the 5-Second Check?
Everyone who pastes into an AI tool. The 5-Second Check is not a leadership job. It is not a compliance job. It is a paste-time reflex owned by the person doing the pasting. Leasing professionals, maintenance techs, marketing coordinators, property managers, regional managers, executives. Everyone. That is why Week 06 taught it to everyone.
Today · Week 07
How an AI Council Actually Works
The operating half of governance. Three flows. One rhythm. Zero mystery.
Skip the warmup anytime. Click Begin Reading when you are ready.
Multifamily NEXT · AI Tuesdays · Week 07
How an AI Council Actually Works
The operating half of governance. Three flows. One rhythm. Zero mystery.
For six weeks you learned where data can go and where it never goes. That is the classification half of governance. This week we teach the operating half. What the AI Council actually does when a request lands on the table. How Tier 2 gets approved without dragging for weeks. How Tier 3 gets refused without shutting down initiative. How new tools get reviewed before anyone uses them.
Twenty minutes. Three flows. When you close this workshop you will know how to run an AI Council, or how to ask your leadership to start one.
Last Tuesday · Week 06
The 5-Second Check
Five things, five seconds, before you paste. That check is your personal reflex. Today we teach the shared reflex that runs at the organization level.
Beginner-friendly · Read at your own pace · Star slides, add notes, print, or email yourself anytime
How This Session Is Organized
What is inside
Twenty minutes of reading. No timer. Pause whenever you need. Every acronym is spelled out on first use, so no prerequisites required.
Part 1 ~3 min
Warmup and opening
Recap Week 6. Set up the operating question: how does a Council actually run?
Part 2 ~3 min
What the AI Council actually is
Seats, cadence, decision authority. The concrete, not the aspirational.
Part 3 ~5 min
The three flows
Tier 2 approval, Tier 3 refusal, new tool review. The three rhythms that make governance real.
Part 4 ~2 min
The one boundary and the log
What gets logged. The default rule that runs everything.
Part 5 ~3 min
Power Preview
A live walkthrough of drafting a strong Tier 2 submission using an RTF prompt.
Part 6 ~4 min
Practice, quiz, close
Spot the Tier scenarios. Five-question quiz. Download the Charter Template. Preview Phase 2.
Objective: By the end of this session, you will know the three flows every AI Council runs, the four questions every Tier 2 request gets reviewed against, and the eight questions every new tool gets checked with. Plus a formal AI Governance Council Charter Template to adopt today.
Section 1 · Why This Week Matters
Every framework tells you to have an AI Council. Almost none tell you how to run one.
That gap is where governance dies. Read a hundred frameworks and none of them tell you what to do in the first fifteen minutes of your first Council meeting. This week does.
The gap in three numbers
13%
of real estate leaders say they could pass an AI governance audit in the next 90 days.
Grant Thornton 2026 AI Impact Survey · multifamily and construction leaders
18%
of organizations across all industries have an enterprise-wide council with authority over AI governance.
McKinsey 2024 State of AI
77%
of organizations say they are working on AI governance. Near 90% for those already using AI.
IAPP AI Governance Profession Report 2025 (with Credo AI)
Read those in order. Nearly everyone knows they need this. Almost no one has it. In real estate specifically, only 13% could defend what they built. That is the gap Week 7 closes.
Three concrete stakes
1
You are already making decisions. The question is whether they are logged.
A regional approves a leasing chatbot pilot. A property manager tells a leasing professional to stop using ChatGPT for renewal notices. A vendor demos a screening AIAIArtificial Intelligence. Any system that produces output by pattern recognition rather than fixed rules. In multifamily this most often means Large Language Models like ChatGPT or Claude, or specialized tools for screening, pricing, and marketing.. Those are all AI Council decisions. Right now they are getting made by whoever is in the room.
2
Fair Housing auditors will ask for the decision log.
If HUDHUD (US DEPARTMENT OF HOUSING AND URBAN DEVELOPMENT)The federal agency that enforces the Fair Housing Act at hud.gov. Multifamily example: HUD investigates a Fair Housing complaint alleging that an AI-generated leasing script systematically discouraged applicants who indicated they had a service animal. The Council needs to review any AI output that could produce disparate impact on a protected class. Anchor: RETTC Pillar 2 (Promote Fairness & Consumer Protection). or a state Attorney General opens a Fair Housing inquiry and AI touched any decision in the resident lifecycle, the first document requested will be your governance record. No log means no defense.
3
Your teams are asking. Give them a place to ask.
Leasing professionals are already trying AI on their own. Without a visible Council with a way to submit an idea, your team members either stop using AI entirely and fall behind, or use it in the shadows and expose you. A functioning Council turns that energy into approved workflows.
Gut check · Where is your property?
Which of the three stakes is most urgent for your property this quarter?
No wrong answer. The Charter Template on Slide 21 addresses all three. Your top answer here will shape how you use the template.
Teaching Slide A · The Concrete Definition
What the AI Council actually is.
The AI Council is the small group at your property, region, or portfolio who own three things: approving new AI use cases, refusing prohibited ones, and reviewing new AI tools before anyone uses them. That is the whole job.
The three ownership items
Approvals. Every Tier 2 use case that involves proprietary data, resident-facing output, or pricing- or screening-adjacent decisions needs Council review before it runs.
Refusals. Every Tier 3 request gets a documented refusal that names why and redirects the requester to an approvable version where possible.
New tool review. Every AI tool that a team member, executive, or vendor wants to introduce goes through Council intake before anyone uses it.
What the AI Council is NOT
Not an ITIT (INFORMATION TECHNOLOGY)The department typically responsible for computer systems, networks, software, and cybersecurity. Multifamily example: your IT lead approves which cloud tools are safe, sets up single sign-on, and responds when a leasing professional cannot log in to the Property Management System. RETTC Pillar 4 (Protect Privacy & Promote Data Integrity) makes IT a core Council voice, but not the only voice. committee. Legal, operations, and property leadership belong at the table alongside IT.
Not a technical committee. Its job is decision authority, not implementation.
Not a bottleneck. If Council cannot consistently meet the Tier 2 turnaround window it committed to, the process is broken.
Not always three people. Small operators run a two-seat Council. Large portfolios run tiered Councils. Right-size to the operation.
Where the Council sits in the org
The Council is the middle layer of a three-tier structure. Executive leadership above sets the risk tolerance. Operational teams below execute the day-to-day. The Council in the middle turns policy into approved workflows.
Strategic Executive Tier. CEO, COO, or Chief Risk Officer holds final accountability for AI risk tolerance, capital allocation, and policy approval. This tier does not vote on every use case. It sets the guardrails the Council operates inside and reports upward to the board.
Governance Steering Tier · The AI Council. Cross-functional body responsible for policy formulation, Tier 2 approvals, Tier 3 refusals, vendor intake, and ongoing risk monitoring. This is where the three flows on Slide 5 happen. This is what the rest of Week 7 teaches.
Operational Execution Tier. Property managers, regional VPs, IT admins, marketing directors. Monitors day-to-day tool performance, logs system errors, escalates concerns to the Council, and runs local human-in-the-loop review on any AI output that touches a resident.
Standard Council seats
Seat
Represents
Typical Role
Governance LeadChair
Owns the decision log and meeting cadence
COO, VP of Operations, or Managing Partner
Operations Voice
Property management and site perspective
Regional Manager or Director of Operations
Compliance Voice
Fair Housing, data privacy, contracts
General Counsel or Compliance Officer
Technology Voice
Tool architecture, security, integration
CIO, Head of IT, or vendor management lead
Practitioner VoiceRotating
The role most affected by the request under review
Leasing, marketing, maintenance, or accounting lead
Right-sizing to your operation
The five-seat structure is the reference model. Actual Council shape scales with your operation. Do not stall a Council because you cannot fill five seats.
Operator Size
Council Shape
Cadence
SmallUnder 500 apartments
2 to 3 seats. Often: managing partner, one operations lead, one outside advisor (legal, tech, or fractional).
Monthly standing meeting. Async channel for routine requests.
Mid500 to 5,000 apartments
4 to 5 seats. Full reference model with rotating practitioner voice.
Every two weeks. Weekly during a heavy rollout period.
Large5,000+ apartments
5+ seats plus tiered sub-councils by function (leasing AI, screening AI, revenue management AI, back-office AI).
Weekly during rollouts. Sub-councils on their own cadence.
Cadence that works
Standing meeting every two weeks (30 minutes for routine requests). Weekly during a heavy rollout.
Asynchronous review channel (email, Slack, Teams) for Tier 2 requests that do not need discussion.
Emergency intake for Tier 3 refusals: any Council seat can issue a refusal within 24 hours, then the Council ratifies at the next meeting.
Localize · Which seat is closest to your job?
Pick the seat you would most likely hold or influence:
Teaching Slide B · The Operating Model
The three flows that make a Council operational.
A Council that cannot run these three flows on a rhythm is a poster on the wall. A Council that can run them consistently is a governance program.
Flow 1 · Approval
Tier 2 use case in, decision out
Trigger: Team member submits Tier 2 use case
Council job: Review against 4 questions. Approve, refine, deny.
Outcome: Approved use case logged with guardrails and review date.
Flow 2 · Refusal
Tier 3 request in, redirect out
Trigger: Team member or vendor brings Tier 3 request
Council job: Confirm Tier 3. Document why. Redirect where possible.
Outcome: Documented refusal in log. Requester coached, not punished.
Flow 3 · New Tool
New AI tool in, intake decision out
Trigger: New AI tool surfaces from any source
Council job: Run the 8-question intake. Approve, pilot, or reject.
Outcome: Tool status logged. Approved tools list updated.
The 3-3 rule for every flow
Every flow has three inputs: the request, the requester, the context (which use case, which data, which residents affected).
Every flow has three outputs: the decision, the reason, the log entry.
Every flow completes in three touches: intake, review, response.
Gut check
Which flow does your property struggle with most right now?
Your answer shapes the rest of the workshop. The next three slides walk each flow in detail. Pay closest attention to the one you struggled with.
Teaching Slide C · Flow 1
How Tier 2 approval actually works.
Step 1 · Intake
A team member submits a Tier 2 request using a standard form (see Slide 21). Required fields: what is the use case, which tier of data it touches, which residents or applicants it affects, which approved tool would run it, what output the team member expects. If the form is not complete, intake is not complete.
Step 2 · Review against the four questions
Every Council reviews every Tier 2 request against the same four questions:
1
Does this touch protected class information under the Fair Housing Act, FCRAFCRA (FAIR CREDIT REPORTING ACT)The federal law regulating how consumer credit and background information can be used, including in applicant screening. Governs accuracy, disclosure, and adverse action notices when someone is denied. Multifamily example: if AI is used to score or summarize an applicant credit report, FCRA rules apply. Miss those rules and you owe the applicant a specific notice with specific language within a specific window. FCRA violations are typical class-action fodder. Anchor: RETTC Pillar 2 (Promote Fairness & Consumer Protection)., or the ADAADA (AMERICANS WITH DISABILITIES ACT)The federal law prohibiting discrimination based on disability. Multifamily example: an AI-generated leasing script that says "no service animals under 25 pounds" violates the ADA (service animals are not weight-limited). Or an AI screening tool that flags an applicant based on their mental health prescription history. Any AI output that could touch disability status, service or emotional support animals, or reasonable accommodations is ADA-adjacent. Anchor: RETTC Pillar 2 (Promote Fairness & Consumer Protection).? If yes, escalate to compliance voice.
2
Does this involve a resident-facing output or a decision that affects a resident (pricing, screening, renewal, notice)? If yes, human review is required on every output.
3
Is the tool an approved enterprise-tier platform under the Data Classification Policy? If no, route to Flow 3 (new tool review) before approving the use case.
4
What is the review cadence for this use case once live? Monthly for high-impact, quarterly for standard. Nothing is approved without a review date attached.
Step 3 · Decide
Approve as submitted · rare. Reserved for use cases that are near-Green.
Approve with guardrails · the norm. Naming specific guardrails (mandatory redaction, mandatory human sign-off, do-not-touch fields) is where Council value lives.
Deny · with a documented reason and, where possible, a suggested rework.
Step 4 · Log
Every decision, approval or denial, goes into the Council decision log. See Slide 9 for what a log entry contains.
The turnaround commitment
Commit to a turnaround your Council can consistently hit, and publish it internally. Governance that is too slow is governance that gets bypassed. If your Council routinely runs longer than the committed window, your team stops submitting and starts working around you.
Teaching Slide D · Flow 2
How Tier 3 refusal actually works.
The wrong way to refuse a Tier 3 request is to punish the requester. The right way is to confirm the refusal, document the reason, and redirect to an approvable version where one exists. A Council that only says no eventually stops receiving requests. A Council that says no with a redirect keeps the pipeline of ideas coming.
Step 1 · Confirm the tier
Not every request that looks Red is actually Red. Someone asking to "run AI on the rent roll" might really mean "help me summarize occupancy trends from an anonymized export." Before refusing, confirm the request is actually Tier 3 by asking: what specific data would be pasted, what specific output is expected, what specific decision would be made from the output.
Step 2 · Name the reason in one sentence
Every refusal names the specific reason. Not "policy says no." A one-sentence reason like "this request would paste screening report content into a consumer AI tool, which is Tier 3 under our Data Classification Policy" is the standard. Vague refusals get re-litigated. Specific refusals close the loop.
Step 3 · Redirect where possible
Most Tier 3 requests have an approvable cousin.
Refused
"Use ChatGPT to identify delinquent residents from the rent roll."
Redirect
Anonymize the rent roll first (strip names, apartment numbers, ledger notes), then use the approved enterprise tool to summarize aging patterns. No resident-identifying data leaves your desk.
Refused
"Have AI draft the eviction filing text."
Redirect
Legal filings stay with counsel. AI can draft a checklist of documents to gather for counsel review. That is Green.
Refused
"Photograph the applicant driver license and let AI extract fields."
Redirect
Never AI-scan identification documents. Manual entry only. If workload is the issue, that is a staffing conversation, not a tool conversation.
Reflect · Your redirect list
Write one Tier 3 request you have already heard at your property. What is the approvable cousin?
✓ Saved locally
Teaching Slide E · Flow 3
How new tool review actually works.
Whenever any new AI tool surfaces from any source: a vendor pitch, a team member finds a tool, an executive reads about one, a conference booth, a competitor mentions one. Before anyone at your property uses that tool, it goes through the intake checklist. This is RETTCRETTC (REAL ESTATE TECHNOLOGY AND TRANSFORMATION CENTER)Industry organization at rettc.org that publishes an 8-point AI Governance Framework written specifically for rental housing. Released November 19, 2025 at OPTECH 2025 in Las Vegas by RETTC's AI Working Group, a group of housing providers and technology partners. RETTC is strategically aligned with the National Multifamily Housing Council (NMHC). Executive Director: Kevin Donnelly. Multifamily example: your property adopts RETTC as the industry standard for AI governance because it names Fair Housing, FCRA, and ADA by hand and gives you a defensible framework in front of an auditor. AI Tuesdays teaches inside the RETTC framework. Pillar 8 (Encourage Third-Party Partner Accountability and Due Diligence) made operational.
The 8-question intake checklist
#
Question
Why It Matters
Blocker?
1. What data does the tool ingest?
If it needs Tier 3 data to work, it cannot be approved.
Yes if Tier 3
2. Where does data go once ingested?
Cloud region, retention period, third-party sharing. In writing.
Yes if opaque
3. Enterprise-tier with a DPADPA (DATA PROCESSING AGREEMENT)A legally binding contract that defines how a vendor handles the personal data you send them. Required by most data privacy laws before you can share resident information with a third-party AI tool. Multifamily example: a leasing chatbot vendor collects prospect names and phone numbers. Before you turn that on, you need a DPA that specifies where the data is stored, how long, who else sees it, and what happens if you cancel. No DPA, no data. Anchor: RETTC Pillar 4 (Protect Privacy & Promote Data Integrity).?
Consumer tools do not qualify.
Yes if no DPA
4. SOC 2 Type 2SOC 2 TYPE 2A standard security attestation for enterprise software vendors, audited by an independent CPA firm. It confirms that a vendor has controls in place around security, availability, processing integrity, confidentiality, and privacy, tested over a period of time (usually 6 to 12 months). Multifamily example: your Property Management System vendor gives you a SOC 2 Type 2 report each year. New AI tools should too. If a vendor cannot provide one, they are not enterprise-ready. Anchor: RETTC Pillar 4 (Protect Privacy & Promote Data Integrity). attestation?
Standard security attestation. Ask for the report.
Yes if none
5. Resident-affecting decisions?
If yes, Fair Housing and FCRA exposure. Compliance voice must review.
Escalation
6. Bias testing documented?
Ask for the audit. If the vendor cannot show one, that is your answer.
Escalation
7. Named owner at your property for output review?
No owner, no approval. Human oversight requires a named human.
Yes if none
8. Exit path documented?
If you cancel, what happens to your data and your workflows?
Yes if none
Three decision paths
Approve for use. All 8 questions cleared. Tool goes on the approved list.
Conditional pilot. Most tools land here. Time-boxed pilot (30 to 90 days) with named scope, guardrails, and a review meeting on the calendar.
Reject. One or more questions cannot be answered acceptably. Rejection goes in the log with the reason. Vendor gets a clear response.
Real-world sidebar
What about the AI you already have running?
Most operators at exec-review scale already have AI-adjacent tools in production: revenue management, applicant screening, leasing chatbots, marketing automation, voice AI. New Councils inherit them.
Do not pause them. Do not ignore them. Run a retroactive intake.
Inventory first. List every AI-adjacent tool live at your property today. Vendor name, tier of data touched, who uses it, decisions it produces or influences.
Run the 8 questions on each, exactly as if it were new. Any tool that cannot answer them acceptably is a candidate for review, renegotiation, or replacement.
Log the retroactive review the same way you would log a new intake. That log is your Council operating history.
Prioritize by risk. Resident-affecting tools go first. Back-office productivity tools can wait 60 days.
Teaching Slide F · The One Boundary + The Log
The one boundary and what gets logged.
The one boundary
The one bright line the AI Council never crosses: no AI tool at your property makes autonomous decisions that affect residents. Approval, denial, pricing, renewal offer, screening outcome, application status, lease termination. Every one of those decisions has a named human on it. AI can prepare, summarize, and recommend. AI never decides.
The decision log
Every Council decision goes in the log. Every one. The log is your evidence of governance if a regulator, auditor, resident, or insurance carrier ever asks.
What each log entry contains
Field
What it holds
Entry number
Sequential. AC-2026-0142 style.
Date decided
Date of the Council meeting or asynchronous decision.
Requester
Name and role of the team member who submitted.
Use case
One paragraph describing what the AI would do.
Data touched
Which tier, which data categories.
Tool proposed
Which platform would run it.
Council decision
Approve, approve with guardrails, deny, conditional pilot, reject.
Guardrails attached
Specific requirements: redaction, human sign-off, do-not-touch fields.
Review date
When Council will re-review. Monthly, quarterly, one-time.
Signatures
Which Council seats concurred. If any dissented, note it.
Retention
Retention is a legal question, not a workshop answer. Common guidance runs three to seven years depending on jurisdiction, insurance carrier requirements, and record type. Confirm with counsel before setting your policy. Whatever window you pick, apply it consistently and version-control the log.
Section Close · The Default Rule
The default rule.
When in doubt, submit to the AI Council.
Do not use. Do not pilot. Do not activate the workflow until the Council has cleared it.
Why this default is stronger than any policy document
It removes the pressure to guess. Nobody is asking any team member to memorize tier boundaries perfectly.
It sends every uncertain case to the group best equipped to handle it.
It builds the decision log automatically, because every uncertain case becomes an entry.
It gives leasing professionals, maintenance techs, and marketing coordinators a place to send their ideas without fear of being told they should have known better.
Synthesis · Three things I will do in the next 7 days
Write three specific actions you will take before next Tuesday.
Examples: talk to my regional about starting a Council, submit my first Tier 2 request, ask my vendor for their DPA and SOC 2 report.
1
2
3
✓ Saved locally
Section 2 · Q & A
Questions people ask.
Tap any question to expand the answer.
Use this workshop to make the case. Slide 21 has the Charter Template. Take it to your supervisor, regional, or executive team and ask for two things: a standing three-person Council (Operations, Compliance, IT) and a two-week meeting cadence. Start there. Councils grow into their shape.
Dissent gets logged. The majority decision holds, but the dissenting seat is named in the log with their reasoning. This protects the Council. If a later review shows the majority was wrong, the dissent record shows the process worked as designed.
The most senior operations leader you have. In a small operator, that may be the managing partner or property owner. In a mid-size, it is often the director of operations or regional manager. Do not wait for a title match. The Council needs someone with real decision authority. Assign it.
Politely and firmly. "Our AI Council reviews every tool before we use it. That is a policy, not a comment on your product. Please share your DPA, SOC 2 report, and bias testing documentation." If the vendor will not provide these, that is your answer.
For any resident-facing decision (pricing, screening, renewal, notice), yes. RETTC Pillar 3 (Uphold Transparency & Explainability) recommends being prepared to disclose AI involvement in resident-affecting decisions. The safe rule: if AI touched a decision that affects a resident, be prepared to say so if asked.
First offense is coaching, not discipline. The team member likely does not know the process. Second offense means the process needs to be more visible. Third offense enters normal HR territory. The Council is a place to send ideas so they get done right, not an enforcement arm.
Your question · anonymous
What question do you still have about running an AI Council?
Sent to Tami · not attributed to you✓ Thanks · question sent
Section 3 · Power Preview · Setup
Watch a Council submission come together.
The prompt structure we use is RTFRTF (ROLE, TASK, FORMAT)The most reliable prompt structure for professional AI output. Role: tell the AI who it is speaking as. Task: name the specific job. Format: name the exact output shape you need. Multifamily example: "You are a multifamily operations analyst (Role). Draft a Tier 2 AI Council submission for using AI to prepare monthly variance narratives (Task). Format as a one-page structured submission with these sections: Use Case, Data Touched, Tool Proposed, Guardrails Requested, Review Cadence (Format)." RTF gets covered in depth in Week 14.: Role, Task, Format. It is the workhorse prompt for professional AI output. This preview shows a leasing coordinator drafting a Tier 2 submission. Run 1 shows what a weak draft looks like. Run 2 shows what a strong RTF-structured draft looks like.
The scenario
A leasing coordinator wants approval to use AI for drafting renewal outreach emails. She needs to submit a proposal to the AI Council. She has 15 minutes between showings.
The RTF prompt template
Copy this template
ROLE: You are a multifamily operations analyst helping a leasing coordinator draft a Tier 2 AI Council submission.
TASK: Draft a submission proposing the use of an approved enterprise AI tool to help write renewal outreach emails. The submission goes to a five-seat Council reviewing on a two-week cadence. Include guardrails, data classifications, and a review date.
FORMAT: One-page structured submission with these sections in this order: (1) Use Case, one paragraph. (2) Data Touched, list of specific data categories with tier for each. (3) Tool Proposed, name the tool. (4) Guardrails Requested, bullet list. (5) Human Review Points, bullet list. (6) Review Cadence, one line. Plain language. No legal jargon. Aim for a Council reviewer to approve or refine inside 15 minutes.
Slide 13 shows what happens if you skip the RTF structure and just ask AI cold. Slide 14 shows the strong version with the full RTF.
Power Preview · Run 1
Run 1: the weak first attempt.
The prompt (no RTF structure)
"Write something for me to send to my AI Council asking to use AI for renewal emails."
The AI output
Dear AI Council,
I would like to request permission to use AI tools to help streamline the process of writing renewal outreach emails to our residents. This will save significant time and improve consistency across communications. AI-powered writing tools have been shown to increase productivity by up to 40 percent in similar contexts, and I believe our team would benefit from this efficiency.
The AI would be used responsibly and with appropriate oversight. We would ensure all outputs are reviewed before sending. Thank you for your consideration.
Sincerely, [Your Name]
Why this fails
No data classification. The Council cannot review what data will be touched.
Made-up statistic. "40 percent productivity" has no source. Councils reject unsupported claims.
No guardrails proposed. "Responsibly" is not a guardrail.
No named tool. Council cannot approve a workflow that has no named platform.
No review cadence. The Council would send it back for basic missing information.
This is what happens when you ask AI cold. Now watch Run 2 with the RTF template.
Power Preview · Run 2
Run 2: the strong RTF version.
The prompt (full RTF)
Full RTF prompt from Slide 12 pasted into the approved enterprise AI tool.
The AI output
1. Use Case
Draft the first-pass version of resident renewal outreach emails using an approved enterprise AI tool. AI produces a draft, human edits and sends. Estimated time savings: 12 minutes per renewal across a 240-apartment portfolio.
2. Data Touched
Property name and address (Tier 1)
Renewal offer terms, current rent (Tier 2, redacted in prompt)
Resident name and apartment number (Tier 2, filled in offline after AI drafts)
3. Tool Proposed
Enterprise ChatGPT (existing DPA and SOC 2 Type 2 on file). Not consumer ChatGPT.
4. Guardrails Requested
No resident names in prompts. Use [RESIDENT NAME] placeholder.
No apartment numbers or ledger data in prompts.
No protected class references (household size, disability accommodations, familial status).
Do not touch: rent amount discussions of Fair Housing categories.
5. Human Review Points
Leasing coordinator reviews every draft before send.
Property manager samples 10 percent of sent emails monthly.
6. Review Cadence
Council reviews use case in 90 days. Then quarterly.
Why this works
Every Council review question is answered before it is asked.
Guardrails are specific and testable.
The tool is named. The DPA and SOC 2 are named.
Human review is named at two levels.
The review cadence is on the calendar before approval.
Same AI. Same 15 minutes. The RTF structure is the entire difference.
Power Preview · Try This Today
Try this today.
Take the RTF prompt template from Slide 12. Substitute your own use case. Run it in your approved enterprise AI tool. Bring the output to your next Council meeting or send it up the chain for review.
The 4-step try-this-today
Step 1. Pick one AI use case you have been holding back because you were not sure if it was allowed.
Step 2. Copy the RTF template from Slide 12. Swap "renewal outreach emails" for your actual use case.
Step 3. Paste into your approved enterprise AI tool (not consumer ChatGPT).
Step 4. Take the output to the next AI Council meeting. If your property does not have a Council yet, take it to your supervisor with a copy of the Council Charter from Slide 16.
If your property does not have a Council yet
Small and mid-sized operators (100 to 2,500 apartments) rarely have a dedicated Chief AI Officer or a data science team. That does not remove the legal exposure under Fair Housing, FCRA, and ADA. It just means you build a smaller version. Here is the workable governance minimum you can put in place this week without a formal Council structure.
1. Named tool owner per AI application. For every AI tool touching your property (leasing chatbot, screening AI, revenue optimizer, marketing generator), name one internal person by title (VP of Operations, IT Director, Regional Manager). That person is accountable if the tool fails or produces a Fair Housing incident. No AI application operates at your property without a named owner.
2. One-page AI inventory. A short register that lists: tool name, named owner, what data goes in, what data comes out, whether the tool produces resident-facing output, whether human approval is required before that output ships, and last review date. One page. Updated quarterly. Available on request.
3. Contractual vendor guardrails. Every vendor contract for an AI tool must include: mandatory notification when the vendor changes the model, on-demand delivery of audit logs, and indemnification if the vendor's algorithm causes a Fair Housing or FCRA violation. If a vendor will not sign to these, that is your answer.
4. Public AI statement on your website. A two-paragraph statement in plain language: what your organization believes about AI use, how you protect resident privacy, and how a resident can reach a human if they think AI has affected them. Two paragraphs. Public. Signed by leadership.
Four items. A one-person Council. This is the minimum defensible position under Fair Housing scrutiny if you cannot yet stand up a full Council. Slide 21 walks the 90-day path to the full version.
Share · Coworker handoff
Copy this message and paste it to a coworker who should see this workshop.
This is how AI Tuesdays spreads inside your property.
Hey. I just did the Week 7 AI Tuesdays workshop from Multifamily NEXT: How an AI Council Actually Works. It walks the three flows (Tier 2 approval, Tier 3 refusal, new tool review) and gives you a formal AI Governance Council Charter Template. Twenty minutes, self-guided, complimentary. Password 0752. Link: https://multifamilynext.com/52-ai-tuesdays/week-07-of-52/
Reference · Bookmark This Slide
The Council Charter · a one-page reference.
A Council Charter is the shortest document you need to start an AI Council. Copy this template, fill in the blanks, adopt it. This is the reference version. The full downloadable version is on Slide 21.
AI Council Charter · [Your Property or Company Name]
Purpose
To approve, refuse, or refine every AI use case at [Property/Company Name]. To review every new AI tool before adoption. To maintain a decision log that documents governance for audit, regulatory, and insurance purposes.
Standing Seats
Governance Lead (Chair): [name and title]
Operations Voice: [name and title]
Compliance Voice: [name and title]
Technology Voice: [name and title]
Practitioner Voice (rotates by use case under review): [pool]
Meeting Cadence
Standing meeting every two Tuesdays at [time]. Asynchronous review channel for routine Tier 2 requests. Emergency Tier 3 refusal authority: any seat can refuse within 24 hours, ratified at next meeting.
The Boundary
No AI tool makes autonomous decisions that affect residents. Approval, denial, pricing, renewal offer, screening outcome, application status, lease termination. Every decision has a named human on it.
Turnaround Commitment
Publish the turnaround windows your Council commits to and can consistently hit. Common shapes: Tier 2 decisions within a defined business-day window of complete intake; Tier 3 refusals within 24 hours; new tool intake within a defined window. Windows scale with operator size and Council capacity.
Decision Log
Maintained by Governance Lead. Retention: 7 years minimum. Each entry contains request, requester, tier, tool, decision, guardrails, review date, signatures.
Anchors
NIST AI RMF (Govern function). RETTC AI Governance Framework Pillars 1 (Establish an Organizational Philosophy on AI Use and Oversight), 3 (Uphold Transparency & Explainability), 5 (Maintain Outcome-Based Accountability & Human Involvement), 8 (Encourage Third-Party Partner Accountability and Due Diligence). Fair Housing Act, Fair Credit Reporting Act, Americans with Disabilities Act.
Reference · Bookmark This Slide
The Vendor Intake Checklist · a one-page reference.
Any new AI tool. Any source. This is what the Council asks before anyone at your property uses it. Full downloadable version on Slide 21.
AI Vendor Intake Checklist
Section 1 · Basics
Vendor name, product name, primary contact.
Sponsor at your property (who is asking to use this).
Use case in one sentence.
Which tier of data does the tool touch (1, 2, or 3)?
Section 2 · Security and Privacy
Data Processing Agreement (DPA) attached.
SOC 2 Type 2 report attached (or explanation why not).
Cloud region where data is stored.
Retention period after cancellation.
Third-party sharing policy.
Section 3 · Fair Housing and Bias
Does the tool touch resident-affecting decisions? If yes, this section is required.
Conditional pilot · 30 to 90 day scope with named guardrails and review meeting on calendar.
Reject · one or more sections cannot be answered acceptably. Reason documented.
Reference · Bookmark This Slide · Standards Library
NIST + RETTC · the anchors under this workshop.
Everything in this workshop maps to two US-focused frameworks: the NIST AI Risk Management Framework and the RETTC AI Governance Framework. Plus the two US enforcement authorities that matter most for multifamily: HUD Fair Housing and FTC.
NIST AI Risk Management Framework · Govern function
How this workshop bridges to NIST: the three Council flows (approval, refusal, new tool review) are exactly what NIST Govern asks organizations to operationalize. Your decision log is your NIST Govern evidence.
RETTC AI Governance Framework · all 8 Pillars
The Real Estate Technology and Transformation Center published the industry-specific 8-pillar AI Governance Framework for rental housing. Reproduced below in its complete published form, verbatim from rettc.org/frameworks/ai-governance:
1. Establish an Organizational Philosophy on AI Use and Oversight
•
Adopt a clear philosophy that defines how AI will be used, governed, and monitored across operations. This organizational approach should balance innovation with accountability, ensure compliance with applicable laws, and promote trust with stakeholders.
Week 7 anchors here: names the AI Council itself.
2. Promote Fairness & Consumer Protection
•
Ensure AI systems comply with applicable provisions of the Fair Housing Act, Fair Credit Reporting Act, the Americans with Disabilities Act, as well as relevant state and local data regulations.
Week 7 anchors here: names the statutory backdrop.
3. Uphold Transparency & Explainability
•
Promote a general understanding of when AI influences key decisions in the customer journey.
•
Provide renters and applicants clear, understandable explanations of decisions affecting them as required by law.
Week 7 anchors here: the resident-disclosure practice (Q5).
4. Protect Privacy & Promote Data Integrity
•
Adhere to federal, state and local consumer privacy laws, with an emphasis on data minimization, security, and informed consent.
Week 7 anchors here: the guardrail patterns in Council approvals.
5. Maintain Outcome-Based Accountability & Human Involvement
•
Engage humans in significant decisions in the customer journey.
•
Establish processes to review model outcomes and provide escalation channels for complaints and appeals.
•
Educate staff on the effective use and limitations of AI tools.
•
Develop testing and ongoing evaluation to ensure AI tools remain aligned with applicable laws, as well as business goals and objectives.
•
Encourage evaluation of AI tools based on their ability to maintain consistent and accurate data flows across the renter journey.
Week 7 anchors here: every decision names a human owner.
6. Promote Positive Renter Experiences and Outcomes
•
Elevate technologies that enhance affordability, access, and housing stability.
•
Explore AI tools that improve resident experiences by streamlining communication, personalizing services, and supporting inclusive resident engagement.
•
Consider engaging renters and resident groups in conversations about how AI enhances their housing experience to build long-term relationship based on open communication and trust.
7. Strive for Innovation with Responsibility
•
Encourage development of AI tools that benefit residents, prospects, guests, on-site teams, and operators, delivering better experiences and more efficient operations, without compromising trust.
•
Educate all stakeholders on how AI enhances renter opportunity and housing affordability, and the role that responsible innovation plays in building thriving communities.
8. Encourage Third-Party Partner Accountability and Due Diligence
•
Promote the responsible development of AI tools across external partners by incorporating robust due diligence and risk management controls. Third-party AI tools, platforms, and data sources are encouraged to align with the organization's established risk and governance practices.
•
Invite transparency from third parties regarding data sources, model limitations, and known risks to support responsible adoption.
Week 7 anchors here: Flow 3 (new tool review).
US enforcement backdrop
Fair Housing (HUD): the Fair Housing Act is enforced by the US Department of Housing and Urban Development at hud.gov. If AI touches any decision that could produce disparate impact on a protected class, HUD exposure is real. RETTC Pillar 2 (Promote Fairness & Consumer Protection) names the specific statutes: Fair Housing Act, Fair Credit Reporting Act (FCRA), Americans with Disabilities Act (ADA).
Federal Trade Commission (FTC): the FTC enforces unfair or deceptive AI practices, including undisclosed AI use in consumer-facing contexts. If your leasing chatbot does not identify as AI to a prospect who asks, that is FTC territory.
Note: AI Tuesdays teaches NIST + RETTC exclusively. US-focused curriculum for US-based multifamily operations.
Section 4 · Practice · Spot the Tier
Practice · 8 scenarios.
Tap Green, Yellow, or Red for each scenario. Feedback shows immediately. No penalty for wrong answers, just coaching. Reset all with the button at the bottom.
1
A leasing professional wants AI to write a friendly welcome email to send to new residents. No names, no apartment numbers, no lease terms.
Correct. No identifying data. Approved enterprise tool. Human sign-off before send. This is exactly the kind of use case that should be running without a Council submission.
Not quite. Nothing here needs Council approval. No identifying data, no resident-affecting decision, no unusual tool. Green.
2
A regional manager wants to paste the current rent roll (names, apartment numbers, ledger notes) into AI to identify aging patterns and delinquency trends.
Correct. The rent roll contains Tier 3 information. Even for internal analysis, pasting resident-identifying financial data into any AI tool crosses the never-share line. Redirect: anonymize the export first, then use an approved enterprise tool.
Not quite. The rent roll is Tier 3. Names plus apartment numbers plus ledger notes plus AI equals Red. Anonymize first, then re-classify.
3
A property manager wants AI to draft variance narratives for the owner report using de-identified GL summary data (no resident names or apartments).
Correct. GL summary data with no resident identifiers is Tier 2. Approved enterprise tool, human review, Council log entry. This is a common Yellow that becomes routine after the first approval.
Not quite. This is Tier 2. Financial summary data without resident identifiers is internal-sensitive, needs approved enterprise tool and Council review before it becomes routine.
4
A maintenance supervisor asks approved enterprise AI to draft a vendor outreach email about preventive HVAC service. No resident data. Vendor pricing terms from a current contract are pasted for context.
Correct. Vendor contracts and pricing are Tier 2 internal sensitive information. Approved tool, but the use case needs Council review of the specific workflow before regular use.
Not quite. Vendor pricing is Tier 2. Even in an approved tool, the specific workflow needs Council log entry before it becomes routine.
5
An executive wants to let an AI screening tool automatically deny applicants below a certain credit score threshold. Fully autonomous, no human review.
Correct. This crosses the one boundary the Council never crosses. AI can prepare, summarize, and recommend. AI never makes autonomous decisions that affect residents. FCRA adverse action notice requirements plus Fair Housing exposure make this a hard refusal with no redirect for autonomous decision authority.
Not quite. This is the boundary from Slide 9. Autonomous resident-affecting AI decisions are Red with no redirect. Human decision authority is not negotiable.
6
A marketing coordinator wants AI to generate photo captions for the property's public Instagram, using property amenity descriptions from your marketing kit.
Correct. Publicly-available marketing content plus an approved enterprise tool plus human review before posting. This is exactly the Green use case that keeps momentum going.
Not quite. Public marketing content is Tier 1. Approved tool. Human review. Nothing here needs Council approval. Green.
7
A property manager wants to use a new AI leasing chatbot vendor she found at a conference. The vendor promises it does not need any review because it uses public data.
Correct. Every new tool goes through Flow 3 intake regardless of vendor claims. Even if the tool ends up being approved, the process is not optional. Request DPA, SOC 2, bias testing.
Not quite. This is Yellow because it is a new tool that needs Flow 3 review. Vendor assurances do not skip Council intake. Even good tools go through the 8 questions.
8
A resident asks the leasing office directly whether AI was used to generate the renewal offer letter they received. The staff member is unsure how to answer.
Correct. Resident-facing disclosure is a Yellow use case. RETTC Pillar 3 (Uphold Transparency & Explainability) says if AI touched a resident-affecting decision, you should be prepared to say so. Council should have a standing disclosure line for these questions.
Not quite. This is Yellow, not Green. Resident-facing AI disclosure needs a Council-approved standard response. Do not improvise. Route to Council for a policy answer.
Section 5 · Take Action · Anonymous Quiz
Five-question quiz.
Anonymous. No score sent unless you submit.
Five questions. Tap an answer to see the coaching. When you are done, you can optionally submit your results anonymously so Tami can see which questions the audience finds hardest.
1
A team member submits a Tier 2 request. The Council reviews it and approves it "with guardrails." What is the correct next step for the requester?
Answer: B
Approval with guardrails means the workflow starts, but the requester confirms the guardrails and review date in writing first. This creates the paper trail. Slide 6 covers the four-step approval flow.
2
Someone at your property proposes letting AI automatically deny leasing applicants who fall below a specific credit score. No human review in the loop. What does the Council do?
Answer: C
This is the one boundary the Council never crosses (see Slide 9). AI can prepare, summarize, and recommend. AI never decides on approvals, denials, pricing, or any action that affects a resident. RETTC Pillar 5 (Maintain Outcome-Based Accountability & Human Involvement) makes this explicit. FCRA adverse action rules are one live area of AI regulation, and Fair Housing exposure is real. Refuse. No redirect available for autonomous decision authority.
3
A vendor tells you their AI screening product does not need Council review because "you already approved the type of tool." What is the correct Council response?
Answer: B
Every new tool. Every source. Every intake. RETTC Pillar 8 (Encourage Third-Party Partner Accountability and Due Diligence) does not have vendor-assurance exceptions. Slide 8 covers the 8-question checklist.
4
A resident calls the leasing office and asks whether AI was used to calculate her renewal offer. What is the Council-aligned response?
Answer: B
RETTC Pillar 3 (Uphold Transparency & Explainability) plus Fair Housing trend line. The Council should have a standing disclosure response so staff do not improvise. Never deny AI involvement if it existed. Never leave the resident without an answer.
5
The Council disagrees on a Tier 2 request. Two seats want to approve. One wants to deny. What is the correct process?
Answer: B
Dissent gets logged, but the majority decision holds. This protects the Council and preserves the value of dissent. Q&A Q2 covers this pattern (Slide 11).
✓ Submitted anonymously. Thank you.
Section 6 · Take Action · Download
Download the AI Governance Council Charter Template.
A formal, board-ready charter template built to Multifamily NEXT's professional standard. Ten seats (eight voting, two rotating advisory). Five commitments. A ten-item operating cadence. A four-tier escalation path. A dedicated Fair Housing accountability statement. Signature block. Bring it to your executive team. Fill in the brackets. Adopt it as your governance document.
The 90-day roadmap
The template is a document. This is what you do with it over the next quarter to make the document real.
Phase 1 · Days 1 to 30
Charter and inventory
Appoint an Executive Sponsor. Establish a small cross-functional Council using the seats on Slide 4. Complete a portfolio-wide audit that catalogues every AI tool in use (internal and vendor-embedded). Publish an Acceptable Use Policy naming Tier 1, Tier 2, and Tier 3.
Phase 2 · Days 31 to 60
Risk classification and vendor controls
Audit vendor contracts. Insert model-change notification, audit log delivery, and indemnification clauses. Classify every AI tool as high, medium, or low risk. Institute human-in-the-loop triggers on any AI touching screening adverse actions or legal communications.
Phase 3 · Days 61 to 90
Audit, verify, publish
Run baseline disparate impact audits on resident screening and dynamic pricing algorithms. Publish a two-paragraph AI transparency statement on your corporate website. Set a quarterly review rhythm to track drift, incidents, and regulatory change.
Week 07 · The Charter Template
AI Governance Council Charter Template
Full charter: purpose, five commitments, 10-seat membership, decision rights, operating cadence, four-tier escalation, Fair Housing accountability, amendment clause, signature block. Vendor intake checklist and decision log included as appendices.
The email button opens your mail client. Fill in your supervisor address before sending.
Guess Green, get seen.·Choose Yellow, stay mellow.·Red? Use your head.
Section 7 · Close · Phase 1 Complete
Session complete.
What you walked away with
The three flows every AI Council runs: approval, refusal, new tool review.
The four questions every Tier 2 request gets reviewed against.
The eight-question vendor intake checklist for every new AI tool.
The one boundary every Council enforces: no autonomous resident-affecting AI decisions.
The decision log fields and 7-year retention rule for Fair Housing evidence.
The formal AI Governance Council Charter Template to adopt at your organization.
Phase 1 is done. Nice work.
Weeks 1 through 7 built your Risk-First Foundation: classification, hallucinations, tier deep dives, Fair Housing, the 5-Second Check, and now the AI Council operating model. You have the reflexes and the governance model. Phase 2 zooms into how AI actually works under the hood.
Next Tuesday · Phase 2 begins
What AI Actually Is
Phase 2 (Weeks 8 through 13) opens the hood. What generative AI actually is. Tokens. Prompts. Specificity. Iteration. Why hallucinations and sycophancy happen at the mechanism level. Every week self-guided, 20 minutes, complimentary.
For Your Board · Copy This Paragraph
"We have adopted a standing AI Council to govern artificial intelligence use across our operation. The Council approves new AI use cases, refuses prohibited ones, and reviews every new AI tool before adoption. Every decision is logged. The one boundary the Council maintains is that no AI tool at our property makes autonomous decisions that affect residents. Every resident-affecting decision has a named human on it. Our framework is anchored in the NIST AI Risk Management Framework and the Real Estate Technology and Transformation Center AI Governance Framework."
Use this in a board packet, shareholder letter, insurance renewal (D&O or E&O), or auditor conversation. Adapt language to your operation.
Tami wants to hear from you.
Which slide did you star? What question is still open? Email tami@mprodigitaledge.com or use the "Email My Notes" button at the top of the workshop.
