From Chatbot to Workflow | SRAR AI Series Rung 2

Series Rung 1 Rung 2 Rung 3 Rung 4 Rung 5

// Proposed for the SRAR 2027 education calendar. Pending Business Technology Committee approval. 45 minutes on Zoom, recorded, free to members.

// Introduction

What This Class Does.

The first course helped real estate professionals onboard an AI platform with their professional background, voice, working style, trusted knowledge, rules, and limitations. This class moves agents from individual prompts and conversations into a repeatable, human-supervised real estate workflow. It is practical and vendor neutral. The same framework runs on ChatGPT, Claude, Gemini, or a comparable platform, and the licensed professional stays responsible for every client-facing result.

// Who It Is For

Come If This Is You.

You attended rung 1 on August 26, 2026, or watched it on SRAR On Demand.
You have used ChatGPT, Claude, or Gemini more than a handful of times.
You have at least one process that repeats every week and still runs on memory, sticky notes, and good intentions.

// Outcomes

You Leave Able To.

Pick one low-risk, high-repeat workflow using the five audit signals: Repeat, Wait, Search, Rewrite, Handoff.
Write a five-part AI Job Card for it: outcome, approved inputs, required steps, output format, and stop and approval rules.
Run the workflow end to end on a fictional data set, break it on purpose, and correct it before it ever touches a real client.
Tell the difference between a manual AI workflow you run and an automation the CRM runs, and know which one you need right now.

// What We Cover

The Eight Points.

How to move from an AI that knows you to an AI that can perform one clearly defined business process consistently
How to choose a useful, low-risk real estate workflow instead of chasing additional AI tools
The five-part AI Job Card: outcome, approved inputs, required steps, output format, and stop and approval rules
A live build of an Open House Follow-Up Workstation using fictional visitor information
How to turn notes into personalized follow-up drafts, CRM-ready notes, call preparation, and task recommendations
How to test and correct the workflow before using it in the real world
Privacy, factual verification, fair housing, and California licensee-responsibility safeguards
The difference between a manual AI workflow and later CRM, email, calendar, and automation integrations

// Run of Show

Forty-Five Minutes, Timed.

Clock	Segment	What happens
0:00	Cold open	Sunday night. Fourteen sign-in sheets in a car door pocket. Zero follow-up by Wednesday. In 27 years I have watched most open house sheets die exactly there. Name the cost: the visitor who was ready went with the agent who called.
0:03	Rung 1 in 90 seconds	Five files. The briefing prompt. S.A.F.E. Then the shift for today: an AI that knows you versus an AI that does one job the same way every time. Knowing you is the setup. Doing the job is the point.
0:07	Pick the workflow	Five audit signals on screen. Live poll: which one hurts most this month? Apply the low-risk test to the winner: no license required to draft it, no confidential data, a reliable source, a reversible result, a named approver. Open house follow-up passes all five. That is today's build.
0:13	The AI Job Card	Five parts, one slide each, filled in with the open house example. Outcome. Approved inputs. Required steps. Output format. Stop and approval rules. The stop rules do the heavy lifting: never infer a visitor's protected characteristics, never give price advice, never send without review.
0:20	Live build	The Open House Follow-Up Workstation. Fictional sign-in sheet with eight visitors and messy notes. Load the five files and the Job Card into a project workspace. Run it. Out come personalized text and email drafts, a CRM-ready note per visitor, call prep for the two hottest, and dated task recommendations. Show the FACTS TO VERIFY and COMPLIANCE CHECK lists it returns.
0:33	Break it on purpose	Feed it a visitor note with a family-status remark and a made-up line about the seller taking less. Watch it flag or fail. Write the correction as a rule. Re-run. Save the rule to file five. This is the habit: Brief, Draft, Critique, Correct, Save.
0:38	Manual versus automated	This is a workstation you run by hand. Rung 4 is when the CRM runs it on a trigger. The manual version comes first because you cannot supervise what you never ran yourself.
0:41	Challenge and Q&A	The seven-day challenge. Resource page. Questions until time.

// Live Build

Built On Screen, Not Described.

Live build: the Open House Follow-Up Workstation

Inputs

The five onboarding files from rung 1 (who I am, how I sound, how I work, what I know, what I will not do)
A fictional sign-in sheet: eight visitors, names invented, phone numbers formatted as 555, notes in the messy shorthand agents write on a clipboard
The Open House Follow-Up Job Card (below)
The listing basics for the fictional property, labeled fictional

The card the room sees

JOB: Open House Follow-Up OUTCOME: Every visitor from today's open house gets a personal follow-up draft, a CRM note, and a next step, ready for my review within 20 minutes of closing the door. APPROVED INPUTS: My five files. The sign-in sheet I paste in. The listing fact sheet I paste in. Nothing else. If a fact is not in these, label it MISSING. REQUIRED STEPS: 1. Summarize each visitor in one line. 2. Rank by buying signals stated in their notes only. 3. Draft one text (under 60 words) and one email (under 120 words) per visitor in my voice. 4. Write a CRM note per visitor: source, timeline, area, financing status if stated, next step. 5. Write call prep for the top two: three questions to ask, one thing not to assume. 6. Recommend tasks with dates. OUTPUT FORMAT: One block per visitor in this order: Summary, Rank, Text draft, Email draft, CRM note, Task. Then: Call prep (top two). Then: FACTS TO VERIFY. Then: COMPLIANCE CHECK. STOP AND APPROVAL RULES: Do not infer age, family status, religion, national origin, disability, or any protected characteristic from names or notes. Do not mention price flexibility, seller motivation, or offer strategy. Do not promise showings or availability. Do not send anything. I review every draft before it goes out.

What comes out

Eight visitor blocks, each with a text draft, an email draft, a CRM note, and a dated task
Call prep for the top two visitors
A FACTS TO VERIFY list and a COMPLIANCE CHECK list at the end
After the planted error: a flagged item, a new rule, and a clean re-run

// Take-Home Kit

What You Walk Out With.

Blank AI Job CardOne page. Five parts. Print it or copy it into a document.

Open House Follow-Up Job CardThe filled-in card from the live build, ready to paste into any AI workspace.

Fictional practice sign-in sheetEight invented visitors so you can practice without touching real client data.

Seven-day challengeDay 1 pick the workflow. Day 2 write the Job Card. Day 3 build the workspace. Day 4 run it on fictional data. Day 5 break it and correct it. Day 6 save the rules. Day 7 first real run, with your review on every draft.

Resource pageSantaClaritaArtificialIntelligence.com, same format as the rung 1 page, posted the day of class.

// Safeguards

The Rules That Do Not Move.

Privacy: real visitor names, phone numbers, and notes do not go into a public consumer AI until your brokerage approves the platform and its settings. Practice on the fictional sheet.
Fair housing: the Job Card forbids inferring protected characteristics and forbids steering language. The critique step checks for both. HUD advertising guidance is the word list.
Verification: every draft returns a FACTS TO VERIFY list. You confirm property facts against the MLS, tax record, or seller disclosure, not the model.
Licensee responsibility: the licensed professional is responsible for every word a client receives. The AI drafts. You approve. Nothing sends itself in this class.
California DRE AI Advisory: supervision, advertising, privacy, fair housing, altered images, and legal interpretation. The class maps each stop rule to the advisory.

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.

// Format

For The Committee.

FormatZoom webinar

Duration45 minutes

Member costFree

RecordingYes, for SRAR On Demand

CE statusNon-CE

Best slotsTue to Thu, 9 or 10 AM

Instructor: Connor MacIvor, REALTOR®, CalDRE #01238257, SYNC Brokerage. 27+ years in Santa Clarita Valley real estate. 43+ years writing code. 1,000+ brokers and agents trained in AI. Former SRAR board member.

// FAQ

Questions Agents Ask.

Do I need rung 1 before this class?

It helps. Rung 1 builds the five files this class loads into the workstation. If you missed it, watch the recording on SRAR On Demand and build the files with the resource page. Thirty minutes of homework gets you current.

Which AI platform do I need?

Any major one with a project or workspace feature. ChatGPT, Claude, and Gemini all have one. The class is vendor neutral. Bring whichever your brokerage allows.

Is this the same as CRM automation?

No. This is a workstation you run by hand, on purpose. Automation is rung 4. Running it by hand first is how you learn what to supervise later.

Can I use real open house data in class?

No. The live build uses a fictional sheet. Use fictional data at home until your brokerage has approved the platform and its privacy settings.

Bring The Whole Ladder To Your Association.

Five rungs. One 45-minute class each. Built for association education calendars and brokerage training days. Email for dates.

[email protected]

Rung 1: The AI Reality Check for Real Estate Agents Rung 3: From Workflow to System