CLOCKWORK · ORGANISED BY DAILY CMO
Hire Your
First AI Employee
One AI worker for your business. Built with you, in one day.
BEFORE THE FIRST SALE
For twenty years, one job never got done.
- ~20 YEARS
Marketing never got done.
- NOV 2023
First time I opened AI.
- 2024
Built my first app with AI.
- DEC 2024
It worked. AI could grow a business.
- JAN 2025
Registered a studio to chase it.
Then I tried to build it. Again, and again.
FOUR TRIES LATER
It began with a friend’s marketing. It ended with mine.
- 2025
Built it twice. Neither stuck.
- LATE 2025
Two bigger swings. Nobody came.
- JAN 2026
Stopped building. Wrote the job for AI workers.
- APR 2026
A stranger paid. Zero words mine.
Marketing was my job that never got done. Yours is probably a different one. Today you hire it.
PROOF IT WORKS
AI made all of this. I just checked it.
I did not write any of it. I set up the workers and checked their work.
HOW AI DID IT
One brand, built by AI, step by step.
- AI found a gap in the market.
- AI built the website and wrote the content.
- AI improved the content using Google data.
- AI tested lead magnets. Visitors gave their email.
- AI built 3 paid offers. Two flopped.
- One worked. AI made it better.
One team did all of it: 23 AI workers, 100+ tools, 60+ skills.
What is next: point the same team at more brands.
WHERE THE WORK HAPPENS
Your AI team works where your business already lives.
Not advice, not a chatbot. The team does real work inside the accounts you already run.
sends and captures email
watches Google search
reads the traffic
takes the sales
runs the website
Every worker reads and writes only what you connect. You see every read and write.
HOW A BUSINESS GROWS
Every brand grows the same five steps.
get found
content · search · adsget the email
lead magnetsbuild trust
emailmake the sale
offerskeep them
repeat buysAI reads the numbers at every step, finds the weakest one, and works there. That is how the brand grows.
WHAT CLOCKWORK RUNS
One team. One schedule.
This is the whole team. Clockwork runs them on a schedule and holds every result for your yes.
WHAT IS CLOCKWORK
I built a system to run them. That system is Clockwork.
It clocks in
It does the job on a schedule, on your own computer.
It waits for you
Every result is a draft. Nothing goes live without your yes.
It is done, your way
You approve. The job is done to your standard.
Friends wanted it too, so I made it a product. You do not hire the whole team today. You hire your first worker, for one job.
NOTHING GOES LIVE TODAY
What you leave with by 5pm:
- One AI worker, set up for your business
- One real draft you check, fix, and approve
- One rule your worker remembers
- Your next run booked, and a time to publish
Everything is a draft until you say yes. Nothing is sent, posted, or paid today.
LIVE DEMO · A REAL BUSINESS IN THIS ROOM
Let's do this on your business.
Question: one real question your customers ask
Your answer: the real answer, in your own words
RUN One question in. One draft out. Nothing is sent.
YOUR FIRST CALL · 3 MINUTES
The draft is back. What do you do?
Every check passes.
The facts are right, but the draft needs a fix.
Something is missing, unsafe, or not part of the job.
Proof: pick one, and give one reason you can point to.
THE BELIEF SHIFT
A clever prompt is not enough.
One prompt gives one answer. A worker gives you the job done every day, and you manage it: clear job, clear limits, check the proof.
ROOM VOTE · CARD 1 OF 6
Who should own this job?
A rambling, non-sensitive customer enquiry arrives. Read it and draft a reply for approval.
Fixed steps
Needs interpretation
Owner-only risk
The job repeats, but each enquiry needs interpretation; the worker drafts and you decide.
ROOM VOTE · CARD 2 OF 6
Who should own this job?
An order ships. Send its tracking link.
Fixed steps
Needs interpretation
Owner-only risk
The trigger, steps, and result stay the same every time.
ROOM VOTE · CARD 3 OF 6
Who should own this job?
Set the price and promise for an unusual customer request.
Fixed steps
Needs interpretation
Owner-only risk
A wrong decision can commit the business and damage trust or margin.
ROOM VOTE · CARD 4 OF 6
Who should own this job?
Approve a quote containing an exception the business has never made before.
Fixed steps
Needs interpretation
Owner-only risk
The precedent and commercial risk belong to the owner.
ROOM VOTE · CARD 5 OF 6
Who should own this job?
Forty public reviews arrive. Find the phrases customers use when they recommend us.
Fixed steps
Needs interpretation
Owner-only risk
The inputs vary, so the worker reads, groups, and gives you a checkable shortlist.
ROOM VOTE · CARD 6 OF 6
Who should own this job?
A booking is tomorrow. Send the standard confirmation.
Fixed steps
Needs interpretation
Owner-only risk
A fixed trigger can run a fixed message with no judgment.
THE WORKER MODEL
Every worker needs three things.
Give it the job
What it does, how, and work you can check.
Set the limits
What it knows, what it can use, when to stop, your yes.
Run the shift
The input, the timing, the history, the proof.
THE BUILDING BLOCK
What makes an AI worker? Seven parts.
- BRAINthe AI model
- JOB DESCRIPTIONits instructions
- HANDSthe tools it can use
- KNOWLEDGEyour business facts
- RAISED HANDstops for your yes
- ALARM CLOCKwhen it wakes
- OUTPUTthe draft it hands you
You write the job description and the knowledge. The rest is built and locked.
THE OPERATING RHYTHM
One rhythm, every shift.
The worker drafts. You decide. Each run builds trust.
Five useful outcomes. One worker you can manage.
Your goal: leave with useful work and a repeatable management habit.
2
Keep the setup. Add what only you know.
The worker is already built. So you spend your time on your facts, your voice, your limits, and your standard.
UNDER THE HOOD
A worker is just a folder of files.
the job: what it does, when it stops
the brain: its role and how it thinks
the step by step how
what goes in, what comes out
what needs your yes
the hands: what it can use
safety tests, so it fails safe
You edit one thing: your business facts and voice. Everything else is built and locked.
WHERE THE LINE SITS
We built the machine. You manage it.
One job · what goes in · what comes out · tools · privacy · stop rules · your yes · backup
Who you serve · offer · voice · facts · examples · limits · call to action · what a good answer looks like · timing
Today you do not touch code, files, tools, IDs, or settings.
YOUR SAVED NOTES
Your saved notes carry your standard into every run.
My worker answers questions about ___ for ___.
A good answer from my business must ___.
It must not make up or change ___.
Stop and ask me when ___.
The worker reads your saved rule on every run. You still check every draft.
PUT YOUR BUSINESS IN WORDS
Fill in four sentences. Do not start from a blank page.
- My worker answers questions about ___ for ___.
- It must never make up or change ___.
- Stop and ask me when ___.
- A good answer from my business must ___.
Do: write, compare, tighten. Done: each blank has a rule you can test.
LIVE SYSTEM WALKTHROUGH
Use the current system in front of you.
Follow the facilitator on the live system.
Find where its input starts and where its held output waits.
Point to the path from input to draft.
USE The labels you can see today, not an old screenshot.
YOUR TURN · 15 MINUTES
Make the Answer Page Writer yours.
- Open your saved notes
- Add your four business rules
- Add one example in your voice, and one call to action
- Save, reopen, and check the text is still there
Done looks like: your text is still there after reopen, and your partner can test each part.
Backup: fill in the printed notes, and use the sample run later.
HARD STOP 11:15
A finished run proves the worker ran. It does not prove the work is right.
IF YOUR LAPTOP ACTS UP
A tech problem will not stop you.
If Clockwork will not run for you, use our ready-made copy and keep going. No one waits.
Keep moving: use the printed workbook if the laptop needs support.
3
Do not ask whether it looks good. Prove whether it is good.
Your first real run counts when you can name one problem, fix it, and prove it changed.
ONE REAL RUN STARTS HERE
Give it one safe question and your own answer.
Question: [a real question your customers ask]
Owner answer: [facts, limits, examples, and next step in your own words]
Do not paste a customer chat. Take out names, contacts, passwords, and private records.
HOW TO CHECK
Check in this order.
A clear answer · good layout · waiting for your yes
Facts · numbers · prices · dates match what you typed
Clear · useful · your voice · one clear point
Four questions: made-up facts? answer first? sounds like you? one clear point?
WHAT CHECK LOOKS LIKE
Check → Name → Correct → Prove → Approve
Run the three layers.
Name one problem you can see.
Write one clear fix note.
Run it again and compare.
Only when every check passes.
- Made-up or changed fact
- Answer buried or unclear
- Wrong voice
- Missing a needed part
- Should have stopped
- Something else
YOUR TURN · THE FULL LOOP
Run it. Find one problem. Prove your fix worked.
- Run your real question
- Check three layers. Name one problem.
- Send back one clear fix
- Rerun, compare, then Approve, Send back, or Stop
Done looks like: before and after proof, one named problem, and a decision you can back up.
Backup: use the sample draft and mark it BACKUP, not your own result.
HARD STOP 12:05
A finished run proves the worker ran. It does not prove the work is right.
NOTE VERSUS RULE
A note fixes one draft. A saved rule guides every run.
“Put the answer in the first two lines.” Fixes this draft.
“Every answer must give the answer in the first two lines.” Used on every run.
A rule guides future runs. You still check every draft.
YOUR TURN · SET THE LUNCH RUN
Save the rule before you line up new work.
- Write and save the rule
- Type a new safe question and your answer
- Set it to run at 12:50. Check it is set.
- Leave Clockwork open and the laptop awake
Done looks like: rule saved, new input, run set, and your guess written down.
Backup: write down the sample run ID and your guess.
HARD STOP 12:30
A finished run proves the worker ran. It does not prove the work is right.
LUNCH SHIFT
12:50. Set, not forgotten.
Rule saved · new question lined up · result waits for you
Clockwork runs while the app is open. Keep the computer awake and locked.
13:30 · RESULT IS WAITING
A timed run is proof to check, not work to trust.
The rule is there. The draft still waits for you.
The old problem is back. Name it and fix the rule.
Nothing ran. Save the status, check the timer, use the sample.
CHECK THE LUNCH RUN · 30 MINUTES
Compare your guess with the result.
- Check if it ran
- Find your rule in the result
- Run all three check layers
- Approve, Send back, or Stop, and save the proof
Done looks like: status, proof of the rule, your decision, and the reason.
Backup: check the sample result and label it honestly.
DEBRIEF 14:05
A finished run proves the worker ran. It does not prove the work is right.
PAIR UP · TAKE TURNS
Show a partner how you ran your worker.
You show the three layers.
You name one problem.
Your partner pushes on the fix.
You show the before and after.
You decide. Your partner signs off.
Done: you defend it, they question it, you decide. Then swap.
A finished run proves the worker ran. It does not prove the work is right.
THE NEXT-RUN HABIT
Every scheduled run gets one review.
Draft ready, stopped, or did not run.
There, true, and right.
Approve, Send back, or Stop.
One line: what you did and why.
THE WORKSHOP WAS WORKER 1
Four workers in 30 days.
Learn the operating loop.
Build a campaign draft.
Build the conversion page.
Build angles and draft ads.
Choose what keeps running.
Each released week: one worker, one real input, one scheduled run, one review call.
AUTOMATABLE WITHOUT FAKE INTEGRATIONS
Every worker runs the same way.
You drop in one safe file.
The scheduled worker claims it.
Source, output, and report wait.
You accept and move the run.
Stopped work keeps its reason.
Nothing sends, publishes, launches, connects, or spends.
THE SAME RHYTHM EACH WEEK
Install it. Run it. Bring the evidence.
- We release one certified worker package
- You install it and add your business context
- You drop one safe input into
new/and set the schedule - You inspect the held output in
review/ - We review the marketing work together
- You give feedback and decide whether to run it again
Bring to each call: run state, decision, one useful result, one friction point.
THE 30-DAY LAB · ORGANISED BY DAILY CMO
The planned lab: three worker calls, then a final review.
Email: one campaign email · Landing Page: one focused conversion page · Ads: three distinct concepts with draft copy
You review every output and transfer approved work into your email, website, or ad tool.
Every package, install guide, safe example, review checklist, call, and support route must pass its release gate first.
Release rule: each package opens only after its install, scheduled run, failure path, and support route are certified.
YOUR 30-DAY WORKER LAB
One worker operated. A four-worker system understood.
Today: answer pages. Next: email, landing pages, and ads—released only when certified, always held for review.