Zensen Media MON · AUG 24 · 1:00–3:00 PM · THE LIBERTY COLLECTIVE
a live, hands-on workshop — laptops open the whole time

Ten steps to
an AI that works

First we build the thing that does the work. Then we point it at the market and get you found. Every step is numbered, and you build all of them today.

STEPS 03–06
Build it.
One folder that knows who you are.
STEPS 07–10
Get found.
Be one of the names an AI gives back.
THREE BUILDS ON THE CLOCK · TEN MINUTES EACH bring the laptop charged — we build on it all afternoon
The plan02 / 36
look up here any time you lose the thread

The whole thing,
in ten steps.

before anything else
Get set up.
01Pick your tool
02Know what you’re driving
the first hour
Build it.
03Build the folder
04Fill it with you
05Teach it a job once
06Let it run without you
the second hour
Get found.
07Find out where you stand
08Learn what it cites
09Write the page
10Build the footprint

The strip along the top of every slide tells you which step you are on. Steps 03 and 09 end with ten minutes where you build, not watch.

Getting oriented03 / 36
ninety seconds, so nobody is lost

Who makes what.

OpenAIChatGPT · Codex
AnthropicClaude · Claude Code
GoogleGemini
Meta · Mistral · DeepSeekthe ones you can download
Model

The engine. GPT-5, Claude, Gemini. Swapping it is like swapping cars — they all drive.

Harness

The app that lets a model actually do things instead of just talk. Claude Code, Codex, Cowork.

Context

Everything it knows in this one conversation. It has a ceiling, and past it, it forgets.

YOU DO NOT PICK A SIDElearn to drive once, then use any of them
STEP 01 / 10  ·  PICK YOUR TOOL04 / 36
STEP 01one brain, several front doors

The app you open is
not the thing answering.

what you see
The app.

A box, a chat history, an upload button. Software — it can be redesigned next month and often is.

what answers
The model.

Running in a datacentre. Identical whether you reach it from the website, the phone or a desktop app.

ChatGPT

OpenAI. The default. Biggest and most general.

Claude

Anthropic. Stronger on long documents and written rules.

Gemini

Google. Already inside Gmail, Docs and Search.

Copilot

Microsoft. Windows and Office, largely the same models.

The app decides what the model can reach. A website can only read what you paste into it. A desktop app can read a folder on your machine — which is the entire difference between steps 01 and 03.

STEP 02 / 10  ·  KNOW WHAT YOU ARE DRIVING05 / 36
STEP 02what is actually happening in there

It predicts the
next word.

It has read an enormous amount of writing, and given some words it works out what words most plausibly come next. Everything else falls out of doing that extremely well.

the consequence
Confident ≠ correct.

Nothing is being looked up. There is no filing cabinet behind it. A fabricated statistic and a real one arrive in the same tone of voice.

where it bites
Specifics.

Names, dates, prices, statutes, quotes — anything about you. The more precise the claim, the harder you check it.

This is called a hallucination, and it is not lying — it has no idea it happened. One habit fixes it: anything stated as fact about your business, your industry or the law gets checked before it leaves the room.

STEP 02 / 10  ·  KNOW WHAT YOU ARE DRIVING06 / 36
the fact that makes steps 03 to 06 necessary

It forgets you
every single time.

inside one chat
It keeps up.

Everything further up the conversation is still in front of it. This is why a long thread gets better as it goes.

a new chat
You are a stranger.

Close the tab, come back tomorrow, and none of it survived. Your name, your business, your standards — gone.

So people re-explain themselves every morning, get a slightly worse answer each time, and quietly decide the tool is overrated. It is not overrated. It has amnesia — and amnesia has a fix.

STEP 02 / 10  ·  KNOW WHAT YOU ARE DRIVING07 / 36
the distinction

Chat stops.
Agents keep going.

chat
Question → answer.

A brilliant consultant. You still go and do the work.

VS
agent
Goal → result.

That same consultant, hired in-house. It breaks the goal into a plan and executes it.

AGENT, FROM AGENCYthe ability to act
STEP 03 / 10  ·  BUILD THE FOLDER08 / 36
STEP 03the mental model

The best hire you
will ever make.

Works until it's done

No clock, no drift, no waiting on Monday.

Takes initiative

Breaks the goal into steps without being asked.

Infinite bandwidth

Ten jobs at once costs the same as one.

About $100 a month

That is the whole salary.

And on day one it knows absolutely nothing about you. Onboard it the way you would a person — context, tools, skills.

STEP 03 / 10  ·  BUILD THE FOLDER09 / 36
the holding company model

The complete starter Agent OS.

OS/
├── CLAUDE.md  and  AGENTS.md
├── README.md
├── context/
│   ├── about-me.md
│   ├── my-values.md
│   ├── my-offer.md
│   ├── ideal-customer-profile.md
│   ├── brand-voice.md
│   ├── business-info.md
│   ├── stack-and-machines.md
│   └── memory.md
├── .claude/
│   ├── skills/[name]/SKILL.md
│   └── agents/[role].md
├── Kabuto Club/
│   └── CLAUDE.md
├── Relvara/
├── Zensen Media/
├── Personal Brand/
├── Real Estate (eXp)/
├── Jarvis/
├── Personal/
└── Active/
Memory is context

Corrections become durable rules in context/memory.md.

Skills are process

Each reusable workflow gets its own folder and SKILL.md.

Pillars add local context

A pillar-level CLAUDE.md stacks with the root instructions.

Start with six, not eight

business-info and stack-and-machines came later, when they were needed. Active/ holds one-offs until they graduate.

STEP 03 / 10  ·  BUILD THE FOLDER10 / 36
the file it always reads

The root file is the front door.

CLAUDE.md or AGENTS.md — loaded before you type a word
one
Point

Tell it where the rest of the context lives.

two
Load

Read only the files this task needs.

three
Ask

If the answer is not there, ask — never invent.

A LEAN FRONT DOORnot a fifty-page encyclopedia
STEP 03 / 10  ·  BUILD THE FOLDER11 / 36
open an empty folder — this is the whole thing

Build the folder.

scan prompt 1 from the handout, paste it whole
01It asks how long you have — Say ten minutes. It will pick its eight best questions.
02Answer honestly, not impressively — Vague answers make vague files. It will push back once.
03Read what it writes — Every file. Correct the first thing that is wrong about you.
04Leave the gaps alone — Everything marked [ASK ME] is your homework, not a failure.
10:00
ON THE CLOCK
you will leave with a v1 — the source spent three hours on his
STEP 04 / 10  ·  FILL IT WITH YOU12 / 36
STEP 04the files you write once

Context belongs in six lean files.

about-me.md

Who you are and what you own.

my-values.md

Principles and non-negotiables.

my-offer.md

What you sell and for whom.

ideal-customer-profile.md

The customer and their actual problem.

brand-voice.md

How you sound — with real examples.

memory.md

Corrections worth carrying forward.

Markdown, not PDFs — it reads them straight through, which is faster and cheaper. These files are assets.

STEP 04 / 10  ·  FILL IT WITH YOU13 / 36
the black box

Own your context.

the chat product's way
It keeps a file on you.

You cannot see it, edit it, or take it anywhere. Ask about a breakup on Monday; find it in your landing page on Friday.

VS
your way
You keep the file.

Plain markdown in your own folder. You read every line, and any harness can pick it up tomorrow.

PORTABLE BY DEFAULTswitch harnesses, keep the brain
STEP 04 / 10  ·  FILL IT WITH YOU14 / 36
the self-improving file

A correction should
survive the session.

you correct it once

“Never make up examples. Use real ones or ask.”

memory.md

It writes the correction back as a short, reusable rule.

next session

It checks the rule before doing the work again.

NOT A TRANSCRIPT DUMPa short list of durable rules
STEP 05 / 10  ·  TEACH IT A JOB ONCE15 / 36
STEP 05three keys worth knowing

How you actually drive it.

/
Find your skills

Type slash and it lists every skill and command you have. This is how you see what your agent can already do.

@
Point at a file

Tag a file or folder and it loads straight in — instead of the agent burning a whole loop hunting for it.

Speak, do not type

Wispr Flow turns talking into clean text anywhere. The source spent three hours talking his context files into existence.

You have eight minutes in a moment. Talking is roughly four times faster than typing — and the interview goes deeper when you are not spelling it out.

STEP 05 / 10  ·  TEACH IT A JOB ONCE16 / 36
holding company → subsidiary

Global and project
context stack.

OS/CLAUDE.md

Who you are, how you work, the standards that never change.

Applies everywhere
+
Kabuto Club/CLAUDE.md

What this one pillar is for — its goals, boundaries and local rules.

Applies inside Kabuto Club
OPEN AT THE TOP, TAG DOWNopen OS → @Kabuto Club → both layers apply
STEP 05 / 10  ·  TEACH IT A JOB ONCE17 / 36
SOPs, for an agent

Skills are the SOPs.

.claude/skills/[name]/SKILL.md
NameWhat the skill is called.
DescriptionWhen to reach for it.
ContentsThe step-by-step, exactly as you would write it for a person.
think of a bookshelf

Only the spines are loaded — every skill's name and description.

It pulls the book off the shelf and reads it only when the job calls for it. A hundred skills, almost no weight.

CONTEXT INFORMS · TOOLS ENABLEskills make it consistent
STEP 06 / 10  ·  LET IT RUN WITHOUT YOU18 / 36
STEP 06what MCP actually is

Your tools all speak
different languages.

Gmail speaks French.
Notion speaks Spanish.
Slack speaks Chinese.
MCP
The translator in the middle.
Your agent speaks English.

It stops being a chat window and becomes the layer that sits on top of every tool you own.

START READ-ONLYwiden the permissions once it has earned them
STEP 06 / 10  ·  LET IT RUN WITHOUT YOU19 / 36
worth doing first

Six connections that change the job.

Apify

A library of scrapers. Instagram, LinkedIn, Maps, Reddit — almost anything public.

Firecrawl

Web search finds a page. This one actually reads it, colours and buttons included.

Composio

One connection carrying all the others — and more than one Gmail.

Chrome DevTools

A clean browser it can drive, so it can check its own work.

Playwright

Your browser, already logged in, for the jobs that need an account.

Higgsfield

Images and video, which the model cannot make on its own.

You never name the tool in the prompt. It looks at what it has and picks — the same way a new hire looks at the apps on their desktop.

STEP 06 / 10  ·  LET IT RUN WITHOUT YOU20 / 36
the part that runs without you

Then give it a schedule.

three times a day
Inbox triage

Read what came in, draft the replies, and flag only what actually needs you.

Start here
every night
The morning report

Pull yesterday's numbers and write the one page you want to wake up to.

Nobody has to remember

Keep it drafting, never sending, until you have watched it do the job right by hand at least once.

It only fires while the machine is awake — which is why people end up putting a spare Mac in the corner. Worth knowing, not worth buying today.

STEP 06 / 10  ·  LET IT RUN WITHOUT YOU21 / 36
you are not managing them — you are enabling them

Stop being the first domino.

one
One agent, one job

You ask for something. It goes and does it.

two
It goes without asking

A trigger or a time sets it off, not you.

three
They hand work along

One finds the thing, another writes it up.

four
You set goals, not tasks

You stop assigning work and start approving it.

once it has your goals and your whole context
“Do smart things.”

Three words. That is the whole prompt.

It can only do that if it knows what you know — and most of that was never written down anywhere. Five minutes a day, dictated into one file. Easiest first proactive job: a watchdog that stays quiet until something looks off.

STEP 07 / 10  ·  FIND OUT WHERE YOU STAND22 / 36
STEP 07say it out loud once — A · E · O

Answer Engine Optimization.

Getting your business named inside the answer an AI gives back — instead of ranking as a link that somebody might click.

what SEO was
Ten blue links.

You showed up on a list. The person scanned it, judged for themselves, and picked. Being on the list was the whole job.

what AEO is
One answer. Six names.

There is no list to scan. It reads everything, decides for them, and answers. If it does not say your name, you were never an option.

It also gets sold to you as GEO, AI SEO, LLMO or generative search. Same thing, four price tags. What it actually asks of you is the rest of this hour.

STEP 07 / 10  ·  FIND OUT WHERE YOU STAND23 / 36
the new shortlist

The list is six names long.

6
names in a typical AI answer

Sometimes three, sometimes ten. Six is the average.

91%
of agents are never named at all

Not ranked low. Not on page two. Absent.

71%
of US metros have no leader yet

Not one name holds even a 15% share. The seat is open.

If you are not one of the six, you did not lose to them. You were never in the room to lose.

Median of 6: Omni Eclipse, 2026 Real Estate AI Search Report, 10 Aug 2026 — 334 answers, 253 queries, 197 markets, Apr–May 2026. · 91% and 71%: FlyDragon 2026 Benchmark — 12,400 AI responses, Apr 2026.
STEP 07 / 10  ·  FIND OUT WHERE YOU STAND24 / 36
this already happened

Your clients are
already asking.

900M+
weekly ChatGPT users
82%
of Americans buying or selling use AI for market information
+527%
growth in AI-referred traffic, year on year

The old game is shrinking, not gone. The new one is already here — and it moves in days, not the years SEO took.

900M+ weekly users: OpenAI, Mar 2026. · 82%: Realtor.com survey of 1,000 US adults buying or selling, fielded Aug 2025. · +527%: Previsible, Jan–May 2024 vs Jan–May 2025.
STEP 08 / 10  ·  LEARN WHAT IT CITES25 / 36
STEP 08where the citations come from

You already control
what it cites.

98.1%

of what ChatGPT cites when explaining why it recommends an agent comes from two places.

Your Google Business Profile

Fill in every single field. Leave nothing blank.

Your own website

Which is what we spend the rest of this session on.

You do not need page one of Google. ChatGPT pulls from position 21 and beyond almost 90% of the time.

98.1% (57.6% Google Business Profile + 40.5% own site, zero to Zillow, Realtor.com, Redfin or Trulia across 1,567 citations): Omni Eclipse, 2026 Real Estate AI Search Report. · Position 21+: Semrush, Jul 2025.
STEP 08 / 10  ·  LEARN WHAT IT CITES26 / 36
two things to get straight before you write anything

Probability,
not position.

the fake promise
“Rank #1 in ChatGPT.”

Nobody can sell you that. The answer changes with who is asking, their history, and their exact words.

the real job
Be understood, then recommended.

Raise the odds you are one of the names it gives — to the customers you can actually help.

ONE
Crawl

It reads your page.

TWO
Index

It files the page away so it can be found again.

THREE
Rank

It decides where you show up when someone asks.

Every AI answer sits on top of those three. If it cannot crawl you, none of the next six slides can help you.

STEP 09 / 10  ·  WRITE THE PAGE27 / 36
STEP 09the content map

Four page types teach
when to choose you.

Best Choice

Answers when you are the right fit.

Common customer questions
Reasons to Choose

Answers why your service is different.

One page per service
Local Service

Answers where you are relevant.

Service × location
Head-to-Head

Answers compare and prefer.

Factual · highest risk
STEP 09 / 10  ·  WRITE THE PAGE28 / 36
extraction before applause

Seven rules make an answer extractable.

01Answer in the first sentence.
02One question per URL.
03Write in third person.
04Use clear lists and headings.
05Write at roughly 7th-grade level.
06Use real third-party proof.
07Keep it short — test 200–300 words.

Length barely moves citation. Speed to the answer does. Never bury it in the middle.

STEP 09 / 10  ·  WRITE THE PAGE29 / 36
best choice · the one to start with

Start with the exact question.

the customer's actual words — this is the page title

Who is the best realtor for first-time buyers in Cincinnati?

01The title is the customer's question.
02The first sentence names the best fit and why.
03List the differentiators in plain language.
04Add one verifiable review or example as proof.
ONE QUESTION · ONE ANSWER · ONE URL
STEP 09 / 10  ·  WRITE THE PAGE30 / 36
now make it write something

Build the writer.

scan prompt 2 — it needs the folder you built back in step 03
01Give it three real questions — Three things a customer actually asked you this month, in their words.
02Let it build the skill — Seven rules, four page types, and a hard rule against inventing anything.
03Run it on your strongest question — Watch a page get written about your business, in your voice.
04Read it like a stranger would — Then tell it the one thing it got wrong, and watch the rule stick.
10:00
ON THE CLOCK
this is the one you show someone tonight
STEP 10 / 10  ·  BUILD THE FOOTPRINT31 / 36
STEP 10cadence after quality

Publish slowly. Learn quickly.

one
Bank

Real questions, services, areas and proof — from calls and emails, not from your imagination.

two
Draft

One original answer per question. Its own URL.

three
Verify

Check every fact. Read every comparison twice.

four
Refresh

Update the useful pages every quarter.

Once, before any of this: add your site to Google Search Console. Free, five minutes, and it is how you raise your hand and say “come read me.” Then five to seven pages a week — start below that until the quality holds. AI cites content roughly 26% fresher than what ranks in ordinary Google.

25.7% fresher (1,064-day median for AI-cited URLs vs 1,432 for Google's top ten): Ahrefs, ~17M citations analysed.
STEP 10 / 10  ·  BUILD THE FOOTPRINT32 / 36
off-page, in priority order

Trust is built off
the page too.

01Bing Places — Feeds ChatGPT's local data. Most people have never claimed it.
02Foursquare — Behind a large share of the local businesses ChatGPT names.
03Google Business Profile — 57.6% of what it cites. Complete every field.
04Apple Business Connect — Maps, Siri, Safari, Spotlight. Free, and relaunched this year.
05Yelp — Licensed 330 million reviews to OpenAI last month. Claim the free profile.
06Everything else — LinkedIn, BBB, Zillow, Angi, Chamber. Same name, address and phone everywhere.

One inconsistent phone number is worse than a missing profile — it tells the engine nobody here is sure who you are.

57.6%: Omni Eclipse, 2026 Real Estate AI Search Report. · Yelp licensed 330M reviews and 8M listings to OpenAI — disclosed in Yelp's Feb 2026 earnings, detailed by Axios Jul 2026.
STEP 10 / 10  ·  BUILD THE FOOTPRINT33 / 36
the two biggest movers

Reviews and YouTube.

reviews are rocket fuel
53.5%
with a profile of just 1–13 reviews
vs
1%
with none

Share of relevant AI answers you appear in. The study measured Trustpilot; that Google and Yelp behave the same way is a fair bet, not a finding.

youtube is the one to start
It reads every word you say.

The whole transcript is fair game, so say the phrase out loud and put it in the title. “Best neighbourhoods in [your city]” is enough to start.

Reads your whole transcript, not just the title
53.5% vs 1%: Seer Interactive, 804,491 AI responses across 1,926 brands, Mar 2026 — measured on Trustpilot profiles specifically.
The capstone34 / 36
both halves become one

You did not write a page.
You built what writes them.

Question bank

Real customer language.

AEO skill

The rules and four templates.

Draft

One answer-first page.

Human verify

Facts, proof, legal risk.

Schedule

Publish only after the manual win.

Context and skill from steps 03 to 06. Question bank from steps 07 to 09. It drafts; you still read every one. Publishing is one line of config away — and it waits until you have read a week of drafts.

Going deeper35 / 36
named, not taught — and not needed today

If you want to go further.

Back it up, or put it online
GitHub · Vercel

Your whole OS is one folder on one laptop. GitHub is undo and a second copy. Vercel is how a thing you built ends up on the internet.

When files stop being enough
Supabase · Postgres

A real database. You want one when several people share the data, or when there is more of it than a person could sit and read.

Give it your own library
RAG · vector search

Point it at hundreds of your documents instead of six context files. Most people are better served by six good files.

Let it run without you
Hermes · OpenClaw · Cowork

More autonomous harnesses, usually on a machine that never sleeps. More rope — and more that can go wrong while nobody is watching.

Run your own model
Ollama · LM Studio · open weights

On your hardware, nothing leaving the building. Slower, free. Fine-tuning is almost never the answer — context and skills are.

Answer your own phone
Retell · Vapi · Synthflow

A voice agent that picks up, qualifies, and books straight into your calendar. Probably the fastest money on this list.

Not one of these is required for what you built today. Each one solves a problem you will feel first — go looking when you feel it, not before.

The close 36 / 36
that is the whole thing

Time for questions.

5:00 PM TODAY Everything lands in your inbox.

Both prompts · this presentation · every tool we named · how to get started, step by step · and your link to try Relvara. Nothing to write down and nothing to sign up for right now — it is all coming to you.

twenty minutes, free, no pitch
Book a call with us.
zensenmedia.com
calendly.com/zensenmedia/free-ai-consultation
ASK ME ANYTHING — NOTHING IS TOO BASIC you are not behind. almost nobody has started.
TALK TO A HUMAN 513‑968‑8900 team@zensenmedia.com
01 / 36
← → move · F full screen · O overview · T timer
Jump to a slide — Esc to close
08:00