- Trigger
- Every weekday at 6:00 AM
- Middle
- Research, analyze, report, flag
- Exit condition
- Under 250 words, 5 items max. Quiet day? “Nothing worth your time today.”
- Iterations
- Up to 3 retries, then it tells you what it couldn’t reach.
Gate Sends nothing
All free classesAI automations
AI automations for real estate agents: How do I automate my recurring real estate tasks with AI?
Write the job as four lines before you open any tool: what starts it, the steps in order, what done looks like and how many tries it gets. Test normal and failed runs manually. Schedule only with the required capability and permission, and keep the send button yours.
Read and practice on this page. Everything on this page is free: the recording, the slides, the prompts, and the workbook. No account needed. The free portal account is for downloading the files.
The slides, all 21.
Historical slides preserve the original class. For current practice, test normal and failed runs before authorizing a supported schedule. Three successful days are not a reason to stop monitoring; keep failure alerts, an owner and periodic review.
01 / 21: Trigger, Middle, Done
Trigger, Middle, Done
Automations that finish the work you are tired of restarting every morning.
Tip: click the slide, then use your arrow keys.
Read the slides as text
01 Trigger, Middle, Done
AI Acceleration
Real Estate
INNOVATION LAB · NORTH · AUGUST 25, 2026
Automations that finish the work you are tired of restarting every morning.
Presented by Build Things That Build Things
02 Leave with something actually running, not a list of ideas.
The four decisions: trigger, middle, exit condition, iterations. The shape behind every automation, in any tool, for the rest of your career.
One automation live on a timer. Built in the room, saved, scheduled, firing tomorrow morning without you.
A second one written and three more specced, using the same four lines with different nouns.
The five rails that make it safe to stop watching. Nothing reaches a client without you pressing send.
How it runs: the problem, the anatomy, then a full hour hands-on. Bring the task you are most tired of redoing.
03 You are the trigger. That is the bottleneck.
Not your tools. Not the model. You, remembering, every single day, forever.
Every recurring task in your business is currently waiting on one thing: you remembering to start it, and having enough left in the tank to finish it.
04 The things you re-decide every single day.
The inbox. You open it, you read the same categories of email, you make the same four decisions, from scratch, every morning.
Keeping up. You know you should know what changed in AI and in your market this week. You mean to. Some weeks you do.
The ads. You check the numbers when you remember, which is usually after the money is already spent.
Content. You start from a blank page every time, so it happens in bursts and then it does not happen at all.
The follow-up. The part you know is worth the most, and the first thing that falls off in a busy week.
None of these are hard. All of them are constant. That combination is exactly what a machine should be holding.
05 The cost is not the minutes. It is the re-deciding.
You pay the startup tax every time. Reloading the context, remembering where you left off, deciding again what "good" looks like.
The standard drifts. What you do carefully on a Tuesday you do sloppily on a Friday, and neither one is written down.
It stops entirely under pressure. The week you most need the follow-up going out is the week you have the least time to send it.
It cannot be handed off. Nothing that lives only in your head can be given to an assistant, a partner, or a machine.
06 What an automation is, and what it is not.
It is a standing instruction, not a one-off prompt. You write it once and it runs on a condition, not on your memory.
It is not autopilot. It does the gathering, the sorting, and the drafting. You keep the judgment and the send button.
It is not set-and-forget on day one. You watch it for a few days, fix the instructions, then stop watching.
It is not a new tool you have to buy. The tools you already pay for do this today, and most of you have never turned it on.
It will be wrong sometimes. That is fine and expected, which is exactly why the rails and the exit condition exist.
07 Stop doing the task. Start describing the conditions.
"The work you do every day is not a to-do item. It is a specification you have never written down."
The move for the next hour
For the rest of this class you are not doing your work. You are writing the conditions under which your work gets done without you.
08 Every automation is four decisions. Not a magic box.
Miss one and it either never starts, never stops, or does the wrong thing forever.
Trigger
Middle
Exit
Iteration
TRIGGER
What starts it
A time, an event, a threshold, or you. If nothing starts it, it is not an automation.
09 The trigger: four kinds, and you already have all four.
Time. Every weekday at 6:00 AM. Every Monday. First of the month. The most common and the most underused.
Event. A new email lands. A form is submitted. A lead comes in. Something happened, so go.
Threshold. Cost per lead goes above your number. A listing sits past its days-on-market line. Only fires when it matters.
Manual. You press the button. Still counts, and it is the right trigger while you are learning to trust it.
Start manual. Promote to time-based once it has been right three times in a row.
10 The middle: research, analyze, report. Then act, only if you allow it.
Research. Go get the raw material. Pull the ad numbers, read the inbox, search the news, read the listings.
Analyze. Turn raw into meaning. What changed, what is off, what matters, what is noise. This is the step worth paying for.
Report. Hand it back in a form you can act on in 60 seconds. Short, ranked, with the recommendation stated.
Act. Sort, label, draft, schedule, adjust. Every action is either pre-approved by you or it waits in a draft.
Name the steps in that order in your instructions. A vague middle produces a vague result, every time.
11 Exit condition and iteration number: where "it went rogue" actually comes from.
Exit condition is the finish line in writing. "Stop when every email is in one of the four buckets." "Stop when the report is under 200 words."
Say what done looks like, not what good looks like. Done is checkable. Good is an argument.
Iteration number caps the retries. "Try three times. If it still fails, stop and tell me exactly what broke."
Always give it a way to fail loudly. An automation that hides its failures is worse than no automation.
Both go in writing, in the instructions. Not in your head. The whole point is that you are not in the room.
12 The five rails. Every automation you build today clears all five.
These do not slow you down. They are what makes it safe to stop watching.
Nothing sends to a client, a lead, or the public without a human pressing send.
No client names, addresses, phone numbers, emails, or financials go into the tool. Redact to placeholders.
Every factual claim it hands me is flagged for verification before it goes anywhere.
It has a written exit condition and a capped number of retries.
It reports its failures out loud, to me, in the same place it reports its wins.
13 One hour. You leave with at least one automation actually running.
Not a plan to build one. A running one, on a schedule, that fires without you tomorrow morning.
Any assistant with a scheduling feature works: ChatGPT, Claude, or Grok. The menu name differs by tool. Look for Tasks, Scheduled Actions, or Routines. Redact before you paste: no client names, addresses, phone numbers, emails, or financials.
14 Write the spec first. Four lines, on paper, before you touch the tool.
Trigger: _____ . When should this start? Pick a real time. "Every weekday at 6:00 AM."
Middle: _____ . The steps in order. Research, analyze, report, and act only if you have allowed it.
Exit condition: _____ . What does done look like, in checkable terms? "Under 200 words, five items maximum."
Iterations: _____ . How many tries before it stops and tells you it failed? Three is a good default.
Pick the task from the list you were honest about earlier. One task. The one that costs you the most.
Nobody opens a tool until all four lines are filled in. The spec is the work. The tool is just where you paste it.
15 The daily briefing: research, analyze, report, on a timer.
Easiest to turn on, hardest to give up once you have it.
Every weekday at 6:00 AM, do this:
1. RESEARCH. Search for what changed in the last 24 hours in (a) AI tools that a real estate agent could actually use, and (b) the residential real estate market for [MY METRO].
2. ANALYZE. Drop anything that is a rumor, an ad, or a repeat of yesterday. Keep only items where something actually happened or a real number moved.
3. REPORT. Give me at most 5 items. For each one: one sentence on what happened, one sentence on why it matters to me specifically, and the source link. Rank them by what matters most to my business.
4. FLAG. Mark any number or claim I would need to verify before I repeat it to a client.
EXIT CONDITION: Stop when the briefing is under 250 words and has 5 items or fewer. If you find nothing worth reporting, say "Nothing worth your time today" and stop. Do not pad it.
ITERATIONS: If a search fails, retry up to 3 times, then tell me exactly what you could not reach.
16 Five questions, out loud, with the person next to you.
Five of five or fix the wording now. It is one sentence, not a rebuild.
Does it have a real trigger, with an actual time on it?
Is the middle written as named steps in order, not one vague sentence?
Is the exit condition something I could check, not something I would argue about?
Does it have a way to tell me it found nothing, and a way to tell me it failed?
Does anything in it reach a client without me pressing send? If yes, take that out.
17 Analyze, sort, act, report. Four buckets, nothing sent.
Review the inbox, then draft, action, archive, or delete. You approve every one.
Every weekday at 7:00 AM, review my inbox from the last 24 hours.
1. ANALYZE. Read every message. Ignore newsletters and automated notifications unless a real number or a deadline changed.
2. SORT. Put every message in exactly one bucket:
DRAFT - needs a reply from me
ACTION - needs me to do something that is not a reply
ARCHIVE - useful later, no response needed
DELETE - no value, safe to remove
3. ACT. For everything in DRAFT, write a reply in my voice, under 150 words, using my writing samples. Leave it as a draft. Do not send anything.
4. REPORT. Give me the counts per bucket, then the ACTION items ranked by deadline, then a one-line note on anything genuinely urgent.
EXIT CONDITION: Stop when every message is in exactly one bucket and every DRAFT has a draft written. Report is under 200 words.
ITERATIONS: Up to 3 attempts per draft. If you cannot write one because you are missing a fact, put it in ACTION and tell me which fact is missing.
18 Same four decisions. Three more jobs you are already doing by hand.
The ad loop. Trigger: Mondays at 8:00. Research: pull last week’s numbers per campaign. Analyze: what moved, what is above my cost-per-lead line. Report: what to pause, what to scale, what to test. Act: stage the change, do not push it live.
Content generation. Trigger: Fridays at 2:00. Research: this week’s market numbers plus my own closings and notes. Analyze: pick the three that a client would actually care about. Report: drafts in my voice, each with a verification flag on every number.
Deep research on a market or a question. Trigger: manual. Research: primary sources only, no aggregators. Analyze: what is corroborated, what is one source, what contradicts. Report: the answer, the confidence, and what would change my mind.
Notice what did not change. Trigger, middle, exit, iterations. Every one of these is the same four lines with different nouns.
Pick ONE. Write the four lines. Build it if you have time, take it home if you do not.
19 The remembering is automated. The judgment and the send button stay yours.
You did not learn a tool this morning. You learned a shape that outlives the tool.
Watch them for a week. Fix the wording where they miss, and the fix is almost always the exit condition. Then stop watching. Every model you will ever use takes the same four decisions.
20 Let it run for a week. Fix one line when it misses.
Do not rebuild it. The fix is almost always the exit condition.
Let the briefing fire every morning this week. The first time the output is wrong or too long, change one sentence in the instructions and let it run again. Once it has been right three days running, stop watching it and build the second one.
21 Trigger. Middle. Exit condition. Iterations.
One automation running, one more specced, and a shape that survives the next tool.
The daily briefing is live on a timer. The inbox pass is written. Three more are specced on paper. Every one of them is the same four lines with different nouns, and every one keeps a human on the send button.
Put one job on the board, step by step.
The order the class builds it in. Write the ticket first, run it by hand, and only then hand it a timer.
Pick one task you keep redoing
Choose one simple task that comes back every day or week, like your inbox, a morning news check or your weekly content. Pick the one that costs you the most. One task, not your whole business.
Decide what starts it
Pick a trigger. It can be a time, like every weekday at 6 AM. It can be an event, like a new lead form. It can be a threshold, like a listing past your days-on-market line. Or you press start yourself.
Write the work as steps in order
Name each step: collect permitted material, check what matters, and return a short report or draft. Sorting, labeling or moving real records also changes an account; define and authorize those exact actions separately.
Write down what done looks like
Give it a finish line you can check, like “under 250 words, five items at most.” Tell it what to say when it finds nothing, so it stops instead of padding the report.
Cap the retries and make failures loud
Tell it how many times to try, then stop and tell you exactly what broke. An automation that fails quietly leaves you waiting on work that never happened.
Run it manually and inspect the failures
Start with a normal input and a missing input. Read the saved result and failure message. Correct the instruction or source that caused the defect, then rerun the relevant check.
Review the schedule and permission
If your selected tool supports the required trigger and sources, ask it to walk through the proposed schedule, access and output. Verify the current provider instructions and authorize the exact recurring action before enabling it. Choose an owner and ongoing review interval.
Review each external action
Practice with fictional or permitted inputs and keep output as a draft. Review the exact content and destination before any authorized send or publication. Propose email moves or deletions for review unless that specific recurring action was authorized. Removing names alone does not make client material safe to upload.
Use the class workbook · Get the supporting files
Read and work through the lesson here. Downloadable files use your free student account.
Time-stamped lines from the class.
Ryan's own words from the recording. Each time opens the class at that moment.
- Watch from 16:22
“this is probably the biggest lesson is that when you are the trigger, you are also the bottleneck”
A repeatable trigger can reduce dependence on remembering the task.
- Watch from 19:24
“an automation is a standing instruction. It is not a one-off prompt.”
You write an automation once, and it runs on its trigger instead of on your memory.
- Watch from 23:50
“There's a time trigger, so 8 a.m., 9:00 a.m., 10 p.m., whatever. There's an event, an email lands”
A trigger can be a set time or something that happens, like a new email landing.
- Watch from 29:05
“you want to make sure that done is actually validatable. So what is it actually producing? Is it giving you a report?”
Done has to be something you can check, like a report that holds each item you asked for.
- Watch from 30:21
“you want to make sure that when it does fail, it fails incredibly loudly. You don't ever want what's called a silent failure.”
When it can't finish, it should stop and tell you why instead of going quiet.
- Watch from 31:49
“this is a sample of a daily briefing automation so every weekday at 6 a.m. research, analyze, report, and then flag.”
Ryan's sample daily briefing has a set time and names its steps in order.
- Watch from 49:25
“So it wants you to use the chat to build your automation. It wants to get the information directly from you.”
In the live demo, the scheduling feature opens a chat that asks you questions and builds the automation from your answers.
- Watch from 56:49
“Hey, I want you to work through my emails is useful, but it's only a start”
“Work through my emails” is a start, and filing each email where it belongs while leaving only action items is an outcome.
- Watch from 1:03:46
“it's not ever deleting anything out. It's just moving them so that if you want to delete them you can.”
Have the automation move email into a folder instead of deleting it, so the final call stays yours.
Where a ticket goes wrong.
- Staying the trigger yourself
- A recurring job can be missed when its only trigger is memory. After testing the work and failure path manually, consider an approved trigger with monitoring.
- Fixing the result, not the instruction
- When a run comes back wrong, fixing that one result only helps today. Fix the instruction that made it, and tell the AI to add the fix to its rules.
- Calling “make it good” a finish line
- “Good” is an argument. “Done” is something you can check, like a word limit and a number of items. Without that, it pads the report or never stops.
- Letting it fail quietly
- If it can't finish and says nothing, you won't know it broke. Cap the tries and have it report what went wrong in the same place it reports its wins.
- Asking it to “work through my emails”
- That's a start, not an outcome. Say what each run should leave you with: a short report, a sorted inbox, or only the important items sent to you.
- Letting it send or delete on its own
- Drafts should wait for you. Keep anything that reaches a client, a lead or the public behind your send button, and have it move mail to a folder you can check.
Where every ticket goes.
A job written as four lines goes to AI on a schedule, comes back as a draft, and waits for your stamp.
- Trigger
- Middle
- Your approval
- Done
Approvedby you
You leave with this ticket running on a timer, the inbox pass written, and three more specced.
The workbook, ticket by ticket.
Trigger, Middle, Done: Automations That Finish the WorkWhat you build: a tested briefing specification, an inbox draft workflow and additional workflow plans; scheduling requires supported features, source access and explicit permission.
The blank ticket.
The Automation Spec Sheet. Four lines. Every automation you will ever build, in any tool, is these four decisions with different nouns. Fill it in before you open any tool.
1TriggerWhat starts it
A time, an event, a threshold, or you pressing the button. If nothing starts it, it isn't an automation.
2MiddleWhat it does
The work, named step by step, in order. Research, analyze, report, and act only where you've allowed it.
3Exit conditionHow it knows it's done
The finish line in checkable terms. Say what done looks like, not what good looks like.
4IterationsHow many tries
A capped number of retries. “Try 3 times. If it still fails, stop and tell me exactly what broke.”
# AUTOMATION SPEC SHEET ## The job What am I tired of redoing? [ONE SENTENCE] ## 1. TRIGGER - what starts it Pick one: time / event / threshold / manual [e.g. Every weekday at 6:00 AM] ## 2. MIDDLE - what it does, named steps in order 1. RESEARCH: [go get what raw material?] 2. ANALYZE: [turn raw into meaning - what gets thrown away?] 3. REPORT: [hand back in what form, ranked how?] 4. ACT (optional): [sort / label / draft / stage - never send] ## 3. EXIT CONDITION - how it knows it is done Write what DONE looks like, in checkable terms. [e.g. Under 250 words, 5 items maximum. If nothing is worth reporting, say so and stop. Do not pad.] ## 4. ITERATIONS - how many tries before it stops [e.g. Retry up to 3 times, then tell me exactly what broke.] ## The five rails - all five or it is not ready - [ ] Nothing reaches a client, lead, or the public without me pressing send - [ ] No client names, addresses, phone numbers, emails, or financials pasted in - [ ] Every factual claim flagged for me to verify - [ ] Written exit condition and capped retries - [ ] It reports its failures out loud, where it reports its wins
Five tickets, ready to copy.
The same four lines with different nouns. Copy the instruction, fill the [BRACKETS] with your own details, and paste it into your assistant's scheduling feature. Every ticket stops at the gate: it drafts, sorts or stages, and you decide what leaves.
Job 01
Schedule example
TriggerWeekdays, 6:00 AM
You need
- Any assistant with a scheduling feature
- A metro or market you care about
The Daily Briefing
research, analyze, report, on a timer
Every weekday at 6:00 AM, do this: 1. RESEARCH. Search for what changed in the last 24 hours in (a) AI tools that a real estate agent could actually use, and (b) the residential real estate market for [MY METRO]. 2. ANALYZE. Drop anything that is a rumor, an ad, or a repeat of yesterday. Keep only items where something actually happened or a real number moved. 3. REPORT. Give me at most 5 items. For each one: one sentence on what happened, one sentence on why it matters to me specifically, and the source link. Rank them by what matters most to my business. 4. FLAG. Mark any number or claim I would need to verify before I repeat it to a client. EXIT CONDITION: Stop when the briefing is under 250 words and has 5 items or fewer. If you find nothing worth reporting, say "Nothing worth your time today" and stop. Do not pad it. ITERATIONS: If a search fails, retry up to 3 times, then tell me exactly what you could not reach.
From classThe line that makes this survive past week one is the one telling it to say nothing worth your time today and stop. Without it you get five items whether or not five things happened, and you stop reading it by Thursday.
The gate
Hands you a report. Sends nothing.
Done looks like
- Under 250 words
- 5 items maximum, ranked
- Source link on every item
- Explicit "nothing worth your time today" when the week is quiet
Your stamp
Job 02
Draft example
TriggerWeekdays, 7:00 AM
You need
- Any assistant that can reach your inbox, or a manual paste-in
- 3 of your own real sent messages, redacted, for voice
The Inbox Pass
analyze, sort, act, report - nothing sent
Every weekday at 7:00 AM, review my inbox from the last 24 hours. 1. ANALYZE. Read every message. Ignore newsletters and automated notifications unless a real number or a deadline changed. 2. SORT. Put every message in exactly one bucket: DRAFT - needs a reply from me ACTION - needs me to do something that is not a reply ARCHIVE - useful later, no response needed REVIEW FOR DISCARD - a person must decide whether anything can be removed 3. ACT. For everything in DRAFT, write a reply in my voice, under 150 words, using my writing samples. Leave it as a draft. Do not send, archive, delete or change any message or record. 4. REPORT. Give me the counts per bucket, then the ACTION items ranked by deadline, then a one-line note on anything genuinely urgent. EXIT CONDITION: Stop when every message is in exactly one bucket and every DRAFT has a draft written. Report is under 200 words. ITERATIONS: Up to 3 attempts per draft. If you cannot write one because you are missing a fact, put it in ACTION and tell me which fact is missing.
From classExactly one bucket is load-bearing. If a message can land in two buckets you have to re-sort the report yourself, and then you have automated nothing.
The gate
Replies wait as drafts. Nothing is sent.
Done looks like
- Exactly one bucket per message
- Every DRAFT has a written draft, unsent
- Report under 200 words
- Missing-fact items move to ACTION, never guessed
Your stamp
Job 03
Specced
TriggerMondays, 8:00 AM
You need
- Your ad platform reporting
- Your cost-per-lead line, written down
The Ad Loop
pull, analyze, recommend, stage
Every Monday at 8:00 AM: 1. RESEARCH. Pull last week's numbers for every active campaign: spend, leads, cost per lead, and the week-over-week change. 2. ANALYZE. Flag every campaign above my cost-per-lead line of [MY NUMBER]. Separate real movement from normal weekly noise. 3. REPORT. Three lists: PAUSE, SCALE, TEST. One sentence of reasoning each, ranked by money at stake. 4. ACT. Stage the changes. Do not push anything live. EXIT CONDITION: Stop when every active campaign appears in exactly one of the three lists. Under 250 words. ITERATIONS: Retry a failed data pull up to 3 times, then report which campaign you could not read.
From classSet your cost-per-lead line before you run this the first time. Without a number in the instruction, analyze has nothing to compare against and you get a summary instead of a recommendation.
The gate
Changes staged. Nothing pushed live.
Done looks like
- Every campaign in exactly one of PAUSE / SCALE / TEST
- Ranked by money at stake
- Changes staged, never pushed live
Your stamp
Job 04
Specced
TriggerFridays, 2:00 PM
You need
- Your own closings and notes from the week
- 3 real sent messages, redacted, for voice
Content Generation
from real numbers, in your voice
Every Friday at 2:00 PM: 1. RESEARCH. Gather this week's market numbers for [MY METRO] plus the closings and notes I logged this week. 2. ANALYZE. Pick the three things an actual client would care about. Drop anything that only interests other agents. 3. REPORT. Draft each one in my voice using my writing samples. Under 150 words each. 4. FLAG. Put a verification marker on every single number so I check it before anything goes out. EXIT CONDITION: Stop at three drafts, each under 150 words, each with its numbers flagged. ITERATIONS: If you cannot source a number, do not estimate it. Drop the item and tell me which number you could not find.
From classDo not estimate is the rail that matters here. A confident wrong number in a market update is the fastest way to lose a client's trust.
The gate
Three drafts, every number flagged for you to check.
Done looks like
- Exactly 3 drafts
- Under 150 words each
- Every number flagged for verification
- Unsourced numbers dropped, never estimated
Your stamp
Job 05
Specced
TriggerManual, when you have a real question
You need
- Any assistant with web search
- A real question you actually have
Deep Research
manual trigger, primary sources only
TRIGGER: manual, when I have a real question. 1. RESEARCH. Answer [MY QUESTION] using primary sources only: company announcements, official data, regulator publications, original filings. No aggregators, no roundup articles. 2. ANALYZE. Separate what is corroborated by two or more independent primary sources, what rests on a single source, and what contradicts. 3. REPORT. The answer, your confidence level, and what evidence would change the answer. EXIT CONDITION: Stop when every claim carries a source and a confidence label. If the honest answer is that the evidence is thin, say that instead of filling the space. ITERATIONS: Up to 3 search rounds, then report what you could not verify.
From classManual on purpose. This one fires when you have a real question, not on a schedule, and the analyze step is what separates it from a search engine.
The gate
An answer with sources and a confidence label. You decide what's next.
Done looks like
- Primary sources only
- Corroborated / single-source / contradicted, separated
- Confidence stated
- What would change my mind, named
Your stamp
Where the board runs.
Running the four-decision spec on a schedule. Look for the scheduling feature in your tool: the menu name differs and moves, so check the vendor help pages.
- Go to chat.openai.com and sign up or log inUse a work email, not a shared family account.
- Check Data Controls before pasting any contact infoSettings → Data Controls → Improve the model for everyone: OFF.
- Go to claude.ai and sign up or log inUse a work email.
- Check Privacy settings before pasting contact infoUse current provider help to review training, retention and sharing controls for your actual account. Turning off training does not mean zero retention or permission to upload client data.
- Go to gemini.google.com/app and sign in with your work Google account
- Check Keep Activity settingReview the current activity, retention and sharing controls for your account in Google help. Account type and settings matter; no setting alone authorizes client-data use.
The dispatcher's rules.
Check these before considering a schedule. Keep an owner, failure alerts and an ongoing review interval; a checklist does not establish unattended reliability.
This week's dispatch log.
TODAY
Save the briefing specification and test one manual run
Reopen the saved draft and verify its sources and output. Schedule only after the checks pass and your tool and permissions support it.
DAY 2
Inspect the first returned briefing before relying on it.
Check the source claims, missing inputs and usefulness. Correct a material error before another run.
DAY 3
Trace one observed failure to its cause and make a focused correction
Check the inputs, instructions, retrieval and acceptance criteria. Retest the same case before generalizing.
DAY 4
Build the inbox pass from the spec sheet
About 15 minutes. Keep it in draft-only mode for at least a week.
DAY 5
Check the drafts it wrote. Are they in your voice?
Use permitted writing samples with identifying details removed. Compare the revision with the intended meaning and facts.
DAY 6
Review the completed runs and choose the monitoring interval
Let it run.
DAY 7
Spec the third one: ad loop, content, or deep research
Four lines on the spec sheet. Build it when you have a quiet evening.
The ticket drawer.
The Automation Cookbook: Four Platforms, Real Recipes
The class came with a reference companion. One rule runs through all of it: automate the preparation, not the execution. Have it draft, reconcile, or recommend, and keep sending, buying, deleting, and publishing behind your approval. What you build with it: a chosen platform and at least one recipe adapted to a job you actually do, using the blank spec sheet.
Open the drawerClose the drawer: 8 recipe tickets from the cookbook
C1 The Unanswered Inquiry Sweep
Reads recent buyer and seller inquiries, finds every one you haven't answered, oldest first, and drafts a useful reply for each. Drafts only.
C2 The Post-Close Loop
Checks your past-client sheet for anniversaries and seasonal windows in the next two weeks and drafts a touch with no ask in it.
C3 The Deadline Watch
Reads your calendar and deal notes and ranks every inspection, appraisal, financing, title, and closing date in the next 72 hours, with the next action for each.
C4 The Weekly Client Update
Drafts one update per active client: what happened, what's next with a date, what you need from them. If nothing moved, it says so. It never manufactures progress.
C5 The Ad Loop
Pulls last week's campaign numbers, sorts every campaign into pause, scale, or test, and stages the changes without pushing anything live.
C6 Daily Market and AI Research
The daily briefing with a stricter rule: primary sources only, a confidence label on every item, and nothing padded.
C7 Listing and Market Content
Drafts content from real numbers and your notes, drops any number it can't source, and runs a Fair Housing pass that describes the property, never the people.
C8 The Wire-Fraud Thread Check
Scans active-transaction email for lookalike sender domains, changed banking details, and new urgency. Read-only. It never confirms wiring instructions or marks anything safe.
Pick one tool, not four.
Use the assistant you already pay for if it has a scheduling feature. Menu names and plan limits change, so when a page like this one disagrees with the product, trust the product. One automation actually running beats four accounts you opened once.
Build on what you own.
Your inbox, your calendar, your Drive, and your own sent mail. Portal leads arrive as email. Contract dates live in documents you have. Past-client history lives in what you've already written.
The prompt pack, all in one.
Five workflows. Every one is the same four decisions with different nouns: trigger, middle (research / analyze / report / act), exit condition, iterations.
Paste into whichever assistant you already pay for. The scheduling feature is called Tasks, Scheduled Actions, or Routines depending on the tool, and the menu moves, so check your vendor's help pages if you cannot find it.
Redact before you paste anything. No client names, addresses, phone numbers, emails, or financials go into these tools. Placeholders work fine.
Its five prompts are the five tickets
Its rules and week one are the dispatcher's rules and the dispatch log.
The remembering is automated. The judgment and the send button stay yours.
The files for your board.
Create a free student portal account to download this class's files.
Automations prompt pack
All five workflows as copy-paste-ready instructions with the trigger, the named middle steps, the exit condition, and the retry cap already written in: daily briefing, inbox pass, ad loop, content generation, and deep research.
Automation Spec Sheet (blank)
The four lines you fill in before you open any tool: trigger, middle, exit condition, iterations, plus the five-rail safety check. Print it and use it for every job you want to hand off.
Class skill card
The class method on one page, written so you can hand it to your AI as instructions.
Your recurring jobs
Automate the recurring job that costs you most.
Trigger, middle, exit condition, iterations. The Agent Constraint Accelerator applies that shape to the job your business can least afford to keep doing by hand, with a written plan and weekly action items for you and for Ryan Wanner.
Called in from the room.
Real questions agents asked in class, with Ryan's answer and the time it was asked.
Does AI need its own login to each of my email accounts?
Asked at 9:25: “on my Mac I've got all of my emails like coming to one source here. Is this separate where you have to log into Gmail”
Ryan said you connect your email to the AI app you use with a connector, also called a plugin. Once Gmail is connected, the AI can reach your email right from its chat window, so you don't have to bounce between apps.
If I already pay for ChatGPT, should I stick with it or pay for another AI?
Asked at 11:34: “So if you already spent the money on chat GPT and now we have Gemini, like would it make sense to stick with chat GPT”
Ryan said the models are all close enough that you can usually get good results from whichever one you have. He switches only when a company releases a much better model. Most real estate work doesn't need the newest one, though a smarter model asks less technical skill of you.
If AI takes over my reminders, how do I make sure I don't miss things?
Asked at 25:20: “if it's being automated for me, I'm not getting my triggers. So, how do I can I trigger AI?”
Ryan said to set up a morning recap automation. It reads your calendar for today and tomorrow plus the email you sent and received, then tells you what changed and what needs doing. He runs a morning and an evening one, so his reminders come from one place.
If I ask ChatGPT to set up my email folders, will it actually do it in Gmail?
Asked at 59:55: “like like the email folders and all that like you tell it tell it to create within chat and it does on the other platform.”
Ryan said yes, once Gmail is connected. In the app's settings you connect the Gmail plugin, and then when you ask ChatGPT to do something in Gmail, it has a tool to take that action there.
If I read my Gmail in Apple's Mail app, do I still connect Gmail itself?
Asked at 1:01:43: “but I think you still have to go through the email, right?”
Ryan said yes, you still go through the Gmail connector. It works a lot like setting up the Mail app, where you sign in to each account once.
Read the edited class transcript.
Automatic captions from the recording, lightly corrected. Stretches with personal details, unsourced numbers, dated news or advice the class has since changed are left out. That's 18 stretches in all. If the recording and the steps above disagree, go with the steps.
Read the transcript, 12 chapters
2:10 Define the automation outcome
So, today we are going to get into um taking the AI from you kind of manually driving it to trying to figure out what things in your life you can automate. And so there's going to be a thinking framework that's kind of required. Um and then there will be like the actual mechanics of how you automate and how you evaluate um the outputs of those automations, right? Cuz one of the things to remember when we automate something is there are going to be edge cases. There are going to be things that happen uh software updates, things that just can knock the automation off of what it's supposed to do, right? And so we want to make sure that we always have a clear picture of what is the target outcome of this automation so that it
doesn't drift and kind of start pulling in other projects or start doing things that you don't want it to do. So we want very clear ideas on our outcomes. So
3:10 Trigger, work and done
this class is titled trigger middle done because basically what we are doing uh with an automation is we are creating what's called a loop. Now the way a loop works is a loop has three parts to it. A loop has the trigger. So what causes the loop to start? So as an example, uh one of the things that I do is I have an email um basically recap in the morning and in the evenings. So the loop is I think it's like 4:00 a.m. and 700 p.m.
my email trigger runs. So, every day at 4:00 a.m. 7:00 p.m. my email trigger runs. And that's what gets us into the loop. Then there's the middle stuff um or the skill or whatever it is the workflow. The middle stuff is the actual work itself. So, when you're doing the work, what is the work that you're doing? Are you someone who is sorting your emails into different folders? Are you someone who once the email is done, you delete it out? Are you someone who archives everything? Like everybody's working style is going to be different, right? So for me, the way I have mine set up is I care about zero inbox. And so I want to get to a zero inbox. The things that are in my inbox, I personally want to have to take action
on. So anything that is already had the action taken gets moved into like an archive folder or something out of the inbox. So the only stuff that I have in my inbox when I log in is you have to pay this, you need to call this person, you need to do this. It's just those emails, right? And so that's the middle stuff is how I sort things, right? And so how you sort things is going to be different than how I do it. And that's this again, that's where the secret sauce comes in with AI. It's what do you have in your head that makes you kind of function, right? So what is your personal email system? And then we need to figure out how to get the AI to learn that email system as closely as it can.
And then another thing to remember is we're always going to be iterating forward, right? We are always going to be looking at this system and saying, "Oh, it got this email wrong. Here's what I need to do to fix it." And usually that's just going to the chat and say, "Hey, this happened. Resolve this in your rules." Right? [laughter]
To figure out Perfect.
What I recommend doing then is I recommend working with the AI to figure out what your natural work style is. So if you're like, I want stuff out of my inbox, a great way to do it is to say, interview me to help me come up with an email capture system that will work for me. Right? Because I took um there's a conventional methodology called getting things done or GTD. Um that's like you have an action folder, you have buckets, you have all these things that require more horsepower than I have cognitively.
And so what you can do is you can basically say like hey uh take getting things done. Take PARA take these organization systems. This is what I like from them. This is what works for me. This is what doesn't work for me. Or like this is when I feel like I'm the most organized. This is how I feel the most organized. help me translate that into an email system, right? Because like some people like I know people that are just like I want my emails chronologically because I know what day it happened and that's how I'm going to solve it. I'm not going to I don't want them by category cuz that just messes me up, right? There are some people who are like I want these by I want every individual sender in their own little
folder. I want um big categories. So I want like marketing in one folder. I want operations. I want finance. however you do it. And something to remember is that it's no longer you personally going through and operating the system. The AI is going to be the thing operating your email system for you if you go this route with it, right? And so if it looks unstructurable to you, that's not necessarily a bad thing because the AI is going to be able to go in and find the emails for you, right? It's one of those reasons why people can now go back to using spreadsheets that have a thousand rows in them because the AI can manage a spreadsheet with a thousand rows in it, right? Humans cannot. Humans look at that and they're like, uh,
was that
can AI unsubscribe from stuff?
Okay. I do that and they still come in. They still come in. I've done it like hundred times and I feel like
most of my inbox is junk stuff. I try to go through and and and I've even written lists of okay, these are the things that I've unsubscribed and then sure enough like a few months later they're all back.
Yep.
I just wasted all that time for nothing. Sometimes I feel like I need to just do a whole new email address and dump all of that. I don't want to do that. Well,
So, on my
Yeah, I have Gmail and I have Comcast.
I noticed the other day on Gmail you Like if you hover over a line on the right now all
done that but it's still it's just still
it's it's just comes back and and is
and another just dumb question but on my Mac I've got all of my emails like coming to one source here. Is this separate where you have to log into Gmail and and have it work through Gmail
So what you would do with this on the front end is you would hook up something called a connector or a plugin. And so it would all work within whatever AI app you use. So if you use ChatGPT, if you use Claude, Gemini, Grok, whichever one you use, uh if you connect the Gmail plugin to that system, it has direct access to your emails from that interface. So you don't have to bounce around at all. You go right from whatever the chat interf interface is.
people to get started, I would recommend ChatGPT or Claude um for the automation side of it. super duper simple. Um, cursor is really easy as well, but cursor is one of those things that is just introducing a new tool. And most people are better off doing what they know and getting the automations there and then introducing a new tool if they get good enough at it.
And then some of these cost money. So if you already spent the money on ChatGPT and now we have Gemini, like would it make sense to stick with ChatGPT versus like paying? I don't know how much Claude is,
right? So, I don't I personally like the intelligence is kind of negligible at this point. And when I say negligible, I mean they're all close enough to be smarter than I'll ever be that you can usually get good results out of whichever one you need. Um,
if you're on ChatGPT, so ChatGPT's released a couple new models. Um the the only necessarily reason I switch between them is because oftentimes one of the companies will release a new model that is substantially better and smarter. Um wi within the real estate space you guys probably don't need the latest smartest like highest end thing cuz you're not like building like physics-based experiments and stuff like that. Um, but it does the the smarter the model, the less technical you have to be to get good results out of it.
Yeah.
Yep.
Yep.
How like Gemini is really good with videos and photos and stuff. So then like chat is like chat's kind of like my hub of like all things and then if I really want to get really
particular about certain things and I'm starting to use Claude. So I don't know just kind of Yeah. No, it is. And that's one of the things. So the the models are all intelligent enough. They're just getting kind of pushed in different directions
15:50 Build repeatable loops
Um, so today we are going to figure out um how to build our automations, how to build those loops. We're going to talk about triggers, middles, exit conditions, and iterations. So, that is literally all a loop is. It's a trigger. What happens in the middle? Did it meet the condition to exit the loop? And then, what do we need to do to So, there are um let's see. So this is probably the biggest lesson is that when you are the trigger, you are also the bottleneck, right? So if the action requires you to be the thing that takes the action, you're now the bottleneck in that action getting taken, right? Because you're the the person that has to take it. And so if we can set it up to that action to be taken on a trigger, you are no longer
the bottleneck. You're now getting the information and then able to utilize it instead of having to be the one that starts to go get the information, right? like it's much easier. Um, so when I was at T-Mobile, we would have to like manually write our numbers down and like do all that kind of stuff. We would have to go to the system, get the to the data source, get the number, take it to a different data source, write it down, go get to a different data source, get the number, write it down, and then give our manager a thesis as to why our sales are either doing really well or really poorly, right? And so all of that takes time away from selling which is not good in a sales role. Um so now all
of that information would get pulled and sent to me either in an email, a text message, a Slack message, whatever it is. So there's no more waiting on me to be the thing that starts that process. The process is going whether I'm a part of it or not. And almost everybody has things in their business like whether you can think of them right off the top of your head or not that are reoccurring that you do every day. Even like like I said, checking email is probably the thing that I do the most every day that takes up the highest majority of my time. Um, and so building something to get a handle on that was probably one of the best best decisions I've made. Um, so, so like as an example, the things that
you have to make decisions on every day. So like inbox, um, keeping up on project management stuff, if you run ads, if you build content, if you do follow-up, all of that stuff should be starting automatically. you may have a role in it like to identify like reading the email before it gets sent to make sure it's high quality and is you know every all the facts are correct. Um you might need to go change an ad, you might need to uh record a piece of content. Whatever it is, um you will still probably have a role to play, but it's no longer everything's waiting on you to start and then do step two and then step three and then step four. You get to come in at step four, do step four, and then go on
about your day as the automation does step five and six and seven and completes itself. So, the biggest cost is not the minutes, it's having to do the stuff every single day to make those decisions to have kind of that same decision fatigue every single day from the same stuff. Um, so an automation is a standing instruction. It is not a one-off prompt. So you write it once and it runs on that those conditions, not your memory. Which is why it's important to update the automation if you get a bad output. You don't just want to fix the output. You want to fix the system that built the output. Um, so it's not autopilot, meaning if something changes, it's not going to if something changes and your automation isn't set up for that change. It's not
going to know. It could break the automation. And so that's one of those things that we look at it. Oh, it failed. Why did it fail? Okay, let's fix it, right? And so just remember that once you build something, it's done for now because you'll likely have to revisit it as things happen. So, um, a lot of people break down their jobs into like, hey, this is my job. They, what they don't break it down into is these are the individual tasks that make up my job. Every time someone asks me, do I need to be worried about AI replacing my job? My response is maybe, right? And the maybe is if you break down your job into the individual tasks because again if you can describe the task well enough the AI can do it right.
And so here is kind of the four decisions that goes into every automation. Its trigger what starts it. The middle what it actually does the actual work itself. The exit how does it know when it's done? Did it create a report? Did it get to the end of the your email list? Did it like what is the thing that it knows, hey, I did this and now I get to be done with my job. Just like a human, right? Like just like a human knows, hey, once this once I get to this completion point, my job is over. And then iteration. So, um, how do you move it forward when you start to discover things that do break the automation? Because things will.
23:48 Different kinds of triggers
So, there are several different kinds of triggers. There's a time trigger, so 8 a.m., 9:00 a.m., 10 p.m., whatever. There's an event, an email lands, um a new entry goes into your CRM, um a text message, um a lead request form gets filled out, anything like that, any type of event threshold. So a threshold as an example is let's say I'm running ads on Facebook and I say if the ads get more expensive than $10 per cost of thousand impressions I want this ad to shut off right that's a threshold or um let's see oh yeah if a listing sits past it stays on market that might trigger an automation like hey we need to repurpose the messaging of this listing or whatever it is right um So threshold is like once something gets past a certain
point, it triggers like if if you picture like a surfer uh going past the buoys and as soon as they go past the buoys that triggers the Coast Guard to like get into action, right? Like that's kind of the threshold and then manual. So the manual is like you go you press the button to get it started and it works through an automated workflow.
One of the things that I think about saying that is that I contrib.
Yeah. Yeah. Yeah. Yeah.
So, if it's being automated for me, I'm not getting my triggers.
So, how do I can I trigger AI? Can AI get triggered like I do?
So, when you say are you like in your inbox when that's happening or like what context are you in? It could be anywhere.
Well, like I put things in my calendar. I put things in my uh emails. I send an email to myself. Yep. I write things myself. I do different things like that. So, that was one of that's my random thoughts that I'm having while you're doing that is if AI is really controlling all of that, I'm going to miss something and it doesn't know how to act on.
Yep.
Yep. And so the thing that I would the way I would kind of fight against that is I would set up an automation that has like a morning recap basically. So it's going to look at your calendar maybe today and tomorrow. It's going to look at all your emails that you've sent as well as what you've received and it's going to basically say hey based on everything that happened yesterday. So, for example, like I have uh an automation that sends me something every morning and all it's doing is it's looking for changes to my project management software and it's looking everywhere. So, it's looking at all of my work stream. So, if I change something in Google Drive, it notices that and it's like, what changed? Did that impact something on the project plan? Um, if I get an email or send an
email, it's like, hey, these are the emails like we need to make an update. And so every morning it's running through my texts, my emails, my calendar, my Slack, my uh Codex work uh workstream, my Claude workstream, my Grockbot, like anything that I kind of touched on my computer, it's going through and being like, how does all of this relate to that? Because the all of these outcomes, if I'm not if I don't have a big goal for my outcomes, the outcomes are all going in different directions, right? And so if I can get the AI to look at, okay, all of these because I've even had the AI tell me like, hey, you spent way too much time in Codex last week. You need to be selling more. Like it's it's like giving me insights because it it also knows
that I like to build. Like I I kind of had like a discussion on it where I was like, "Hey, help me analyze like what I need to do to move my business forward." And it was like, "You default to building. So if you ever default to building, you need to think about selling in instead." basically. And so it checks all of that and then it sends me the recap in the morning. And so that's where I would really push to get your reminders in. That way your reminders are all coming from one source. You no longer have all of these open flags in your brain. That's like, oh, when I get into this now everything now like alarm bells are going off every time I go like because I'm the same way.
Like I used to go into my email and it would just be alarm bell alarm bell alarm bell. Then I would go to my social media alarm bell and it's like I have all these independent things I need to do that are all a part of my general context, but they're all independent and my brain marks them as independent. So my brain doesn't remind me until I log into the platform. Oh no, I have 50 things that are on fire. I need to do something about that. And so having that morning and like I do an evening report as well. I do a morning and an evening to like recap what happened like what did I do today and then tomorrow the morning is like what do I need to do today? Um and so th that as a system
would probably make a big impact there.
Yeah.
28:56 Exit conditions and retry limits
So exit conditions um so the exit condition is basically when is the automation done getting the result that you wanted it to get. So um you want to make sure that done is actually validatable. So what is it actually producing? Is it giving you a report? And if it's giving you a report, what does that report need to have? Like it needs to have my Gmail recap. It needs to have the uh the Facebook ad spend that we did last week. It needs to have X, Y, and Z, right? And so it's kind of what uh what does that need? And then so the the question then becomes well what if the exit condition never gets met right and it just stays in that loop forever like something breaks in the condition and it just never gets met. So
that is why you put ideally a hard cap on your iterations of the automation. So if you say, "Hey, you can do this. You have five tries to get this done." It's going to try to do it five times. And if it never meets the exit condition, on the fifth one, it'll still exit and it'll tell you why it was not able to meet the exit condition. Maybe there's just information doesn't exist that you think does that should be a part of it. Or maybe the information does exist and it just couldn't find it or something like that. And so it'll just exit and be like, "Hey, this failed. Here's why."
Um, and you want to make sure that when it does fail, it fails incredibly loudly. You don't ever want what's called a silent failure. Um, so as an example, one of kind of the classic user experience problems that people run into, if you've ever had an app and you press a button and nothing happens and you have no idea whether you're waiting on the internet server, you have no idea if it the button just didn't work. Like you have no idea. So if you press a button and nothing happens, that's usually bad user experience. So if you press a button, it should like ding or something to register that the touch is is there. And so if nothing happens, you have no idea that it failed, right? And you're just kind of sitting there waiting.
31:09 Write the spec: a daily briefing
So now we are going to get into actually building this. And so when you're thinking about automations, what you want to think about is with AI uh in general, the planning phase and the quality assurance phase are the two most important phases that a human is going to be a part of. And so for this, we want to think about what does our project spec look like, right? And we're breaking it down into something much smaller than what the average project is, right? So if I want a um and actually I think so it's trigger middle exit condition how many iterations and so this is okay so this is a sample of a daily briefing automation so every weekday at 6 a.m. research, analyze, report, and then flag. So, it's going to research what changed in the last 24
hours. Um, AI tools that real estate a so like as an example, I always do like research for my class. And so, this is uh research AI tools that real estate agent could actually use. Um, analyze, so drop anything that is a rumor report. Give me at most five items for each one. One sentence on what happened, why it matters to me specifically, all that kind of stuff. Um, and so this one is for email. This one is, "Every weekday at 7 a.m., review my inbox from the last 24 hours.
Analyze. Read every message. Ignore newsletters and automated notifications unless a real number or deadline changed. Sort it. Draft anything that needs a reply. Um, take action." So if something needs to be like pushed into your CRM or like into Google Sheets, stuff like that, it can take the action. It can archive the old stuff and it can delete the stuff that has no value to you. Um, and then it can go and act. So it'll draft, it'll actually do the things that are in that automation. Um, it did the research on either the problem, the person, whatever it was that that email was in reference to. It actually went through, found chain of communication, did research, and like is actually getting things information from my uh data to where it needs to go. It's not just sitting there and kind of as a
chatbot anymore. It's actually more of an employee. Um, and so
34:02 Ads and content on a schedule
there are, again, these are really, really good things to use this for. Um, if you run ads, it can do a really good job of evaluating your metrics on your ads and giving you uh the ability to turn on, turn off, make adjustments to your adset, content generation, it's super duper good for. So whether you're looking to do um like you actually record content or whether you want to do like market recaps, you want to do listing stuff, you want to do um just general stuff on the town, stuff about why moving to Tennessee is great, why moving to Hendersonville, whatever it is, um you can build content and you can have it generate automatically every single week. So you don't have to and you can also have it generate ideas and scripts. Um, and I think the content
generation, like the end-to-end thing is something that we're actually going to do in here, like actually build that system so that you don't have to because content, uh, the things that I've seen that are like the big things that realtors would benefit from but don't want to do, content is near the top of the list because it's it's not hard to take your camera and like do the thing, like actually record the content. But what's hard is the editing, the coming up with the idea to to go shoot the thing, coming up with the posting schedule, your analyt, like all that stuff is like it's all solved for, but nobody knows it yet. Nobody knows it yet, especially in the real estate space. So like if
And then
do that same. I have so many videos. I caught driving and fields like so many.
Yeah. And that's honestly that's the best thing cuz if you just put all that content in one Google Drive folder, you can have it generate renewable assets for you over and over and over again. Especially if it's like stock like a lot of people are like, "What am I going to do with this footage of like me driving down the road or of me walking through this house or I mean it right now that's considered like stock footage, right?" And so you just put it in there and it can be the background to your your list.
It can be a part of your talking head video. It's a transition. It can be it's it's basically gives you a full production studio so you don't have to be the creative director of your own business at this point,
Oh, and that's perfect. Yeah. And
that would be good.
Yeah. It it'll be a full like content production studio that literally is I put this in Google Drive, it produces stuff and then I deny and tell it why I denied and then I approve stuff and that's it. Yeah.
Show me your wave. [laughter]
But yeah, so that's Oops. Okay.
38:59 Automate the remembering, keep the judgment
The biggest goal is to take your task from I have to remember to do this kind of what you were talking about to the remembering is done for you and the only thing that needs to happen is the judgment on the outcome like you you're basically trying to position yourself as the AI created this and then you're you know King Solomon or Caesar or whatever and you're yes and knowing everything right or you're saying like maybe I will go in and change these couple things and now it's it's a useful asset. Right. If
Quick question about that.
I'm sure that it can do the same thing with maybe photos in your phone, too, as far as emails, too. So, I have probably go back years now that I have taken tons of videos around Tennessee of all the things. Is there an easier way for me to say get videos in here that are like obviously not videos of my family doing crazy stuff, but is there a faster way than me going back in my camera roll and scrolling and and going through or on the video side? I mean, maybe it is. I spend four or five hours just going through videos to put in, but is there
a way? Yeah.
Faster. So, I believe
that maybe it pulls more of the things I would want and maybe some that I don't want, then it's faster for me to take the ones I don't want.
Yeah.
Video says where it's at.
Yeah. Yeah, it should. Y
I have 106,000 pictures.
That's awesome.
Videos I think is even
Can you search? Can you go to a section that's
Yeah, I can go to my videos. So though I know there's a ton and maybe I can and I can go through and kind of but that was just another question when you're talking about that putting that stuff in a Google Drive. Is there an easier way to do that versus doing it?
So what I would do because I've not actually done this yet, but this is actually a really fascinating uh kind of project. What I would do is I would tell the AI and and I'm actually going to do this so I will have an answer for you next week. So motivation to come. Um [snorts] [laughter] um the uh what I would do is I would say hey uh I would talk to the AI and I would say hey I have all these photos in my photos app that could be beneficial to my real estate business. Please go back and analyze from the beginning of my photo library all photos and videos.
And you could ask it to tag them maybe as like real estate professional. And then that way it would have like a tag that next time you would just be able to be like, "Hey, go pull all my photos that are tagged this and find ones that show me on the lake or that show me in a house or whatever.
Yeah. Yeah. Exactly. Yep. Yep. Yep.
Okay.
Kind of scary thinking access to the phone. I know.
I pulled up
I have everything Google.
Yep.
Um and so I was in my videos it was like ask Gemini. So now I'm like all right what if I ask Gemini to pull all my like landscape looking at houses things and see if he'll because I was just wondering how am I going to put it into Google Drive. So I was just curious.
Yep.
Okay. Sorry. I just
No, that's great.
Hand in hand. I think with a little bit what you're saying because I wondered how the fastest way.
No.
Oh, I do have a lot of good stuff, but oh my gosh.
And so that the that whole line of thinking is exactly why it's important to have roughly the same problem solving process across the board, right? Because with AI um you can solve your finance, you can solve operations, you can solve marketing, you can solve your IT problems, but you just have to have a method to get to the AI to help you solve the problem. Right?
And always when I take pictures, I think that I need to do better each month following them and
a thousand%. No,
no, no one does. And that's why the AI is highly useful. Like I would I would legit have that conversation and say like how do you like help me map this out? Because the the second that you're able to solve one problem on your own. All you have to do then is pattern match to the other problems and you can be like oh shoot I don't know how to do this for accounting. Okay, I'll take it to the AI and I will ask in my business this is what I have. How do I do this particular accounting thing? Right? And the same with marketing, like, oh, I want to send this person a landing page, but I have no web design experience. How do I build a landing page? Go to the AI and have it build your landing p. Right?
And so, like all of the problems, you can enter the problem solving cycle at roughly the same space.
I had I went through my whole pantry and my fridge and my freezer the other day and just had them sitting there on the counter while I talked about this is what I have at two chicken.
Yep.
And and he made like recipes for me for the next three weeks. He goes, "You don't even have you just you go to the store for three things and here's all the stuff that you already have. I mean, it's just like mindset.
Oh, 100%. Yeah.
So, yeah. All this in my freezer. Everyone in this room has something that if you applied to like with full effort, you would have a disproportionate impact on that something compared to anyone else who came into your business and did that something, right? And so the AI allows you to get as much time as possible on your disproportionate impactful thing, right? And so that's what the AI should really be targeted at is like, how do I get back to my most valuable thing? because it can it can
Changed my life in this last week where I say when I wake up and I say these are the things I really need to get accomplished. these are things I wish I could do too and these are the things I really have goals for and this is what I want to make for dinner tonight. Can you help me plan my day? And it is giving me time blocks where I'm staying within these boundaries where where I normally feel so spread thin where I'm just like all over the place like swirling it like crazy because it's too much and too pull too many pulls. It's just saying no don't do this today do this tomorrow.
You're going to be here tomorrow and that will make more sense. I'm like, "Oh my gosh."
Yep.
Yeah.
A game changer for me.
Shocking. It's like personal assistant.
So, this is my personal resource management app. And this is why I think everyone needs uh resource management app. So, this is the resources that I have uh for utilities. So, this is like prompts, this is businesses, this is deals, this is like everything. And so the reason I have this is because when the AI it's kind of like what you were just saying when the AI needs information I it knows where to go to get it. Like if I need someone if I'm like I need someone who is like a Bitcoin cryptographer D. It's going to search all the people in my resource galaxy and it's going to find anyone that may meet that profile or that may know someone that could introduce me to someone with that profile. And so like that's like this is
that thousand-line spreadsheet, right? Like this is what AI is really good at that I'm terrible at is interpersonal connections. [laughter]
You're going to have to quit saying that.
Think that I really like. [laughter]
Well, you guys are first, so you guys you guys are weird, not me.
[laughter]
Yes.
48:39 Set up a scheduled task live
And so like that is so I'm just going to go through I've got ChatGPT and I've got Claude up. And so in ChatGPT it's scheduled. In Claude it is routines. Okay. They're both relatively the same. What you can do is you can click create. And if you just click create, you can see it opens up a chat window and it says, "Let's set up a scheduled task together. First, explain how scheduled tasks work in ChatGPT. Then interview me to figure out what I need scheduled and when it should run." Right? So when I run that and this will if you press create in uh Codex, this is what pops up. So it wants you to use the chat to build your automation. It wants to get the information directly from you. And
so I'm gonna run this.
Yeah. So this is this is the desktop.
Are you on the desktop app or the
is that your mail from from your MacBook?
This
Oh, this is the Codex app. So this is the ChatGPT desktop app.
Oh,
yeah. and I think it might be able to be run from
there's two section
so you can run it from here too. Okay. Yeah. So this is this is the the web version and you can still schedule
I have an app on my computer. Oh yeah, that is
wondering how he integrated it into the mail but it's not the desktop.
Yeah. Yep. So you can schedule it requires more um more work to schedule
the things
just when you click on that they're like do you want us to pay attention like how do you know [laughter]
55:54 Choose the output you actually need
But yeah, honestly, and so you can see here it's so I had it set up a scheduled task. I ran that prompt and it says scheduled tasks are saved prompts that ChatGPT runs later, even when you're offline. They can be onetime reminders, recurring stuff, monitoring tasks, all that kind of stuff. And so now it's going to interview me and it said, "What should the task actually accomplish every time it runs? Do you? And so like as an example, this is really important because you can say should the task accomplish generating an output. So like is it going to send me a Slack? So like as an example for the emails, right?
Like if it's reviewing your emails, do you want it to take all the actions to get you to zero inbox? Do you want it to just generate a report about what's going on in your email address? Do you want it to just send you the important things to a Slack message? Right? Those are three different outcomes that this can get. And so you have to figure out what the outcome is you really want. So like saying like, "Hey, I want you to work through my emails is useful, but it's only a start, right?" So saying like, "Hey, I need you to help me come up with an organizational system for my emails that you'll run every single day, filing things into the folders they belong, and leaving action items in my inbox." Right? Like that is an outcome.
I don't see that when I pull up my chat on on here.
So if you go schedule
Okay.
Yep. And so from there,
what you'd have to say is like, "Hey, interview me to help me create an automated task." And there it'll interview you. If you download the desktop app, it'll be able to go right from um like you'll be able to click a button and it'll just start the interview process. this.
Yes.
So I there's a place I can download it.
Yeah. So there's Yeah, there is a desktop app um that it offers a little bit more as far as like connectivity and your ability to maneuver around your own computer.
58:00 Automate one task this week
But yeah, the the biggest thing that I would tell you guys is figure out one task that you can automate and try to automate it and then come in with it next week and we'll go through it and resolve it if it it needs resolving or we can talk about how wonderful it is or how terrible it is or any. Um, but I would say try to grab a task like just one simple easy task and try to that you is is a time suck for you and try to automate it as much as you can.
So when I try to download um or like do the app, it says AI chat box. It's not that.
No, no, no, no.
Because when I went to the down here to the app store and came up and just put ChatGPT
Okay. Yeah. Go to um Google Chrome.
Oh, I thought I was
Yeah, cuz they've got a weird um the way they have their thing set up. So, yeah, type a ChatGPT desktop app and it should take you to an OpenAI domain. [snorts] Yeah.
Okay,
that's it.
Okay, thank you.
Yep. And so that's pretty much all I had today. Any questions?
That's awesome.
Good.
Yeah. [laughter]
So, was it super intimidating and
No.
Perfect.
Because you're so good. [laughter]
That must be
must be all stink there.
Yeah. Yeah. So if the these are these are cloud run so they run persistently in the background as long as your ChatGPT subscription is live.
59:55 Connecting an email account
Okay. So like like the email folders and all that like you tell it tell it to create within chat and it does on the other platform.
Yeah. So, what you would do is you would set up. So, what you would do is you would go down and you would go into settings and you're going to click plugins and then you you can see Gmail. So, I have a Gmail plugin and so when I tell it in ChatGPT to um do something in Gmail, it has a tool to take that action. And is that like if that's what we want to do this week,
how do we know how to do the plugins?
Yeah.
Will it walk you through it?
Yeah. So, it'll kind of walk you through it. Um
like when you're signing figuring putting your email on your desktop and you have to have certain
Yeah, exactly. So, and I'll give you like a quick super quick tutorial. So, on the desktop app, if you click your name and then click settings and then click plugins. So, it's name, settings, plugins, and then browse directory, and then you'll have all of your apps.
I don't use the Gmail app. I just use
Okay. Yeah. Let me see if it Do you have a Gmail account that you use the mail app?
Yeah.
To the mail because that's what I use, too. app.
Yeah.
I use different a bunch of different
but I think you still have to go through the email, right?
Yeah. So, you'd still go through the Gmail connector. So it it's it's actually very similar to how you would have set up your mail app in Apple, right? Where you go in and you kind of lo you do all the authentication for each one into the mail app.
So can you do that? Because I was realizing that so my ChatGPT is like my besides my email,
but I have like a million other emails. And so I realized that I when I was trying to add the plugin, it was asking me which Gmail account and I'm like I have four or five or whatever.
Yep.
So like how do I
You would a each each one individually.
You can do that. Yeah.
My subscription is under one. So, I'm also a zero notification uh email person.
I just delete them throughout the day.
Yep.
So, I al I almost don't want to give them the option because I might like miss something.
Yeah. No, 100%. And you can also something that you could do is you could say um you could have it just move all the files in like all the emails into like a today folder so it moves them all in. That way they're it's not ever deleting anything out. It's just moving them so that if you want to delete them you can. Or if you want to do whatever you can
Yeah. I I mean I have a bunch of emails that come through
Yeah. Yep.
Yeah.
Oh my gosh.
Yes.
Which the ones that come in. I think I signed up for a lot of stuff on in the email I did not. And then random agents send me their newsletters.
[laughter]
I think you should probably.