Agentic Real Estate | SRAR AI Series Rung 5

Full Series Rung 1: The AI Reality Check Rung 2: From Chatbot to Workflow Rung 3: From Workflow to System Rung 4: The Supervised Automation Stack Rung 5: Agentic Real Estate

TL;DR SUMMARY

An agent uses tools: it reads files, browses, drafts across apps, and runs multi-step tasks. That is the risk chatbots never carried.
The Permission Ladder: read-only, draft-only, act-with-approval, act-within-limits. Nothing client-facing runs unsupervised.
Build a private brokerage intelligence layer, with training disabled and retention set, before any agent connects to real files.
The Transaction Desk turns a fictional escrow file into a task list, a seller update, and calendar reminders, approved at every act.
Audit it monthly and follow a 12-month plan. Act-within-limits is earned, never assumed.

// Introduction

What This Class Does.

The first four rungs taught agents to onboard an AI, give it a job, connect the jobs, and put them on a trigger. This class covers the layer that arrived in force in 2025 and 2026: AI agents that do not just write, they act. They read a folder, browse for a source, draft across email and calendar, and run a multi-step task. Attendees learn the four-rung Permission Ladder that decides what an agent may do on its own, build a private brokerage intelligence layer, and watch a fictional escrow file become a task list, a seller update, a deadline flag, and calendar reminders, with approval at every act. Then they leave with a monthly audit and a 12-month plan for a solo agent, a team, and a brokerage.

// Who It Is For

Come If This Is You.

You completed rungs 2 and 4, or you run an approval-gated automation today.
You are comfortable with a CRM and a shared drive.
You lead a team, manage an office, or carry enough volume that a transaction desk would change your week.

// Outcomes

You Leave Able To.

Explain what an AI agent is (it uses tools: reads files, browses, drafts across apps, runs multi-step tasks) and how that changes the risk compared to a chatbot.
Apply the Permission Ladder: read-only, draft-only, act-with-approval, act-within-limits. Place any real estate task on the right rung.
Build a private brokerage intelligence layer: policies, forms guidance, local knowledge, approved sources, and prohibited items, in a workspace with training disabled and retention set.
Run the Transaction Desk: a fictional escrow file becomes a task list, a seller update, a contingency deadline flag, and calendar reminders, with approval at each act.
Audit it monthly (hours recovered, errors caught, permissions reviewed) and build a 12-month AI operating plan.

// Live Build

Built On Screen, Not Described.

The Transaction Desk

Inputs: the five onboarding files from rung 1 and the brokerage intelligence layer, a fictional escrow folder (purchase agreement summary, six invented emails, an inspection report cover page, the acceptance date), the agent job card below, and a connected calendar in a test account.

AGENT JOB: Transaction Desk PERMISSIONS: Read-only on the escrow folder. Draft-only for any message. Act-with-approval for calendar entries and reminders. No sending. No signing. No contact with any party. STEP 1 (read-only): Read the folder. Summarize the transaction in ten lines: parties by role only, acceptance date, price as stated, contingencies and their day counts, open items. STEP 2 (read-only): Build the task list keyed to the acceptance date. One row per deadline: task, due date, who owns it, source document. STEP 3 (draft-only): Draft the seller update in my voice using only the folder. Label anything not in the folder MISSING. STEP 4 (act-with-approval): Propose calendar entries for each contingency deadline with a reminder two business days prior. Ask me to approve each one. Add only what I approve. STEP 5 (log): List every file you read, every draft you produced, every action you proposed, and every approval you received. NEVER: Give legal or tax advice. Interpret contract language beyond quoting it. Contact any party. Move, delete, or share any file. Act on anything outside this folder.

What comes out: a ten-line transaction summary and dated task list produced read-only, a seller update draft with MISSING labels where the folder was silent, calendar proposals approved one at a time on screen, and an action log of every file read, draft produced, action proposed, and approval received.

The Permission Ladder

Rung	What the agent may do	Real estate examples	Who may run it
1. Read-only	Read, summarize, compare, find.	Summarize the escrow folder. Find every deadline. Compare two disclosure versions.	Anyone on the team
2. Draft-only	Write, but never send.	Seller update. Repair request response for review. Buyer-agent FAQ.	Anyone on the team
3. Act-with-approval	Propose an action and wait for a click.	Calendar entry for a contingency. A reminder. A CRM task.	Licensed agent
4. Act-within-limits	Act without a click, inside written limits, on low-risk sends only.	Monday showing report to the seller if nothing changed and the checklist passed.	Licensed agent, by written exception, after one clean month at rung 3

Never, on any rung: sign anything, send an offer or contract document, give legal or tax advice, contact any party about money, or move, delete, or share a client file outside the approved workspace.

Earning a rung: a task moves up one rung after a documented month with zero rejected outputs at its current rung, and only with the approver's written sign-off. A task moves down a rung the day it produces an error. No exceptions.

Six questions for any vendor

Ask in writing. Keep the answers in the brokerage layer. If a vendor cannot answer all six, client files do not go in.

Encryption

Data encrypted in transit and at rest. They can name the standard.

Retention

You can set how long conversations and files are kept, and delete on demand. They can show you where.

Training on our data

Your data is not used to train their models, and this is the default or a setting you control. In writing.

Admin controls

An administrator can manage users, roles, and what is connected, and can remove a user in one step.

Audit logs

Actions taken by agents are logged with timestamps and are exportable. The broker can read them.

Data location

They can tell you where the data lives and under which jurisdiction.

// Seven-Day Challenge

Under Thirty Minutes A Day.

Read-only first. Always.

Day 1: place five weekly tasks on the Permission Ladder.
Day 2: ask your platform vendor the six questions, in writing.
Day 3: build the brokerage layer starter: policies, one forms guide, one local knowledge page with sources.
Day 4: run the Transaction Desk on the fictional folder, read-only.
Day 5: run the draft-only step. Read every MISSING.
Day 6: run act-with-approval on a test calendar. Approve one. Reject one. Read the log.
Day 7: write your 12-month plan. Put the monthly audit on the calendar.

The 12-month AI operating plan

Quarter	Solo agent	Team	Brokerage
Q1	Five files. One Job Card. Seven-day challenge done.	Every agent has five files and one Job Card. Shared template.	Rung 1 and 2 training for all. Platform chosen with the six questions answered.
Q2	Listing system: fact sheet, chain, critique, template project.	One listing template for the team. Coordinator runs the chain.	Brokerage intelligence layer started: policies, forms guidance, local knowledge.
Q3	One automation live, gate on, red-teamed, measured.	Speed-to-lead for every agent. One after-hours line for the team.	Automation approval checklist adopted. Broker sign-off process live.
Q4	Agent on read-only and draft-only. One act-with-approval task earned.	Transaction desk for the coordinator at rung 1 and 2. Agents at rung 3.	Permission ladder policy adopted. Audit logs reviewed monthly. First act-within-limits exception granted in writing.

// Safeguards

The Rules That Do Not Move.

Confidentiality: escrow files contain protected information. They go only into an approved private workspace with training disabled and retention set, never into a public consumer AI.
Permissions: nothing client-facing runs unsupervised. Act-within-limits is earned after a documented month of act-with-approval with no errors, and only for low-risk sends.
No unauthorized practice of law: the agent quotes contract language and dates. It does not interpret. Legal questions go to the broker or counsel.
Broker supervision: an agent running under a licensee's login is that licensee's activity and the broker's supervisory responsibility. Audit logs exist so supervision is possible.
California DRE AI Advisory: supervision, advertising, privacy, fair housing, altered images, and legal interpretation. The Permission Ladder maps to each.
Access hygiene: before connecting an agent to email or calendar, switch off what it does not need, use a dedicated account where possible, and review the log weekly for the first month.

Every rung of the series runs the S.A.F.E. gate before anything leaves the office. Sensitive information removed. Accuracy confirmed through reliable sources. Fair housing and fiduciary responsibilities protected. Everything properly disclosed and approved.

// FAQ

Questions Agents Ask.

AGENT

Is this safe for a small brokerage?

Safer than what most small brokerages do now, which is agents pasting client data into free chat tools. The private layer, the permission ladder, and the audit log give the broker visibility they do not have today.

AGENT

Do I need to be technical?

No. The class runs the build on screen with plain-language cards. If you can set up a shared drive folder, you can do this.

AGENT

What does an agent cost to run?

Enterprise and team tiers on the major platforms are priced per seat per month. The class does not sell a platform. It gives you the six vendor questions and the permission ladder so the choice is yours.

AGENT

How long until the agent runs on its own?

It does not. Act-within-limits is the ceiling, and it is earned one low-risk task at a time after a clean month of approvals. The licensee stays in command. That is the design, not a limitation.

THE RUNG 5 DECK

See The Slides.

The complete rung 5 slide deck lives on connorwithhonor.com. Open it in a new tab.

Open the rung 5 deck

Resources To Keep Working.

FULL CLASS OUTLINE

Rung 5 class page

Run of show, format, and instructor details on connorwithhonor.com.

Open the class page

PRINTABLE

Rung 5 take-home kit

The same kit content above, formatted for printing.

Open the printable kit

PREVIOUS RUNG

Rung 4: The Supervised Automation Stack

The approval-gated automation loop this class extends into agents.

Open Rung 4

BACK TO THE START

Rung 1: The AI Reality Check

The five-file onboarding kit the whole ladder builds on.

Open Rung 1

BRING THE WORKFLOW. LEAVE WITH THE FIRST MOVE.

Book a free 30-minute AI architecture conversation.

Connor will send a short questionnaire before the call. Bring the task that repeats, the process that stalls, the system that leaks money, or the online identity that refuses to line up. The answer may be automation, coaching, a Meta campaign, MachineFound, a private brokerage system, a better template, or deleting a broken step. The diagnosis comes first.

When appropriate, Connor is comfortable discussing a mutual NDA before reviewing proprietary workflows or business ideas.

Schedule at BookWithHonor.com Text AI to 661-476-2217

QR code to schedule with Connor MacIvor at BookWithHonor.com

SCAN WITH YOUR PHONE
BOOKWITHHONOR.COM

Connor MacIvor · AI Architect · CalDRE #01238257 · SYNC Brokerage · SantaClaritaArtificialIntelligence.com

Educational information only. Not legal, financial, tax, cybersecurity, fair-housing, or transaction-specific advice. Verify requirements with the responsible broker and qualified professionals. Class date: TBD, pending SRAR Business Technology Committee.