After the first useful prompt, the next win is simple: stop
starting over. Give AI one repeat job, one rules sheet, one saved
checklist, and one proof step before you trust the answer.
Random chat is fine once. A workspace is how you reuse it.
A random chat is where most people start. They ask one question,
get one answer, close the tab, and forget what worked. That is
okay for a test. It is not how you build a habit.
Random chat
One-off questions, missing context, repeated setup, and no
saved proof of what actually helped.
AI workspace
One job, saved rules, repeatable steps, and a clear check
before anything gets used.
Start with one workspace for one job. Do not build the whole system first.
Step 2
Pick a job that repeats.
The right first workspace is not fancy. It is the task you keep
doing again and again. If it repeats, AI can help you make the
first draft, checklist, or cleanup faster.
1
Customer replies
Turn messy customer messages into calm first drafts.
2
Weekly notes
Turn scattered notes into tasks, follow-ups, and next steps.
3
Content ideas
Turn one topic into post ideas, talking points, and outlines.
4
Simple research
Collect questions, compare options, and mark what needs checking.
5
Admin cleanup
Turn rough lists into clean checklists a person can follow.
Step 3
Give the bot a rules sheet.
The rules sheet tells AI how to act inside that one job. It is
not a magic spell. It is the same thing you would tell a helper
before letting them touch customer-facing work.
This chat is for one job: [name the repeat job].
My role:
[who you are]
Audience:
[who the output is for]
Voice:
plain, helpful, direct, and not too salesy
Rules:
- ask one question if important details are missing
- do not invent facts, prices, dates, links, or promises
- keep private details out unless I say they are needed
- mark anything I should verify before using
- give me a short version first unless I ask for more
Good example:
[paste a short example of the kind of output you like]
Bad example:
[paste or describe what you do not want]
Job
One workspace should have one main job so the context stays clean.
Voice
Voice keeps the answer from sounding stiff, fake, or too formal.
Rules
Rules stop the tool from acting like it knows things it does not know.
Examples
Examples teach the style faster than a paragraph of instructions.
Step 4
Turn the rules into a workspace card.
A workspace card is the simple version of an AI system. It tells
you what job the workspace does, what goes in, what comes out, and
how you check it before using it.
# AI Workspace Card
Workspace name:
[customer reply / weekly notes / content ideas / research cleanup]
One job:
[what this workspace helps with]
Inputs I provide:
[messages, notes, rough draft, question, screenshots described in words]
Output I want:
[email, checklist, table, script, plan, questions]
Rules:
- ask if important details are missing
- keep the answer plain and useful
- do not invent facts, prices, dates, links, or promises
- mark what needs human review
Proof step:
Before I use the answer, list what changed, what was guessed, and what I should verify.
Starter card
Customer reply
Input is the customer message. Output is a calm first draft and one missing-detail question.
Starter card
Weekly notes
Input is messy notes. Output is tasks, follow-ups, owners, and what still needs a decision.
Starter card
Video idea
Input is one rough topic. Output is a hook, three talking points, and a plain-spoken close.
Step 5
Build the workspace like a small workbench.
The workspace is not supposed to feel like a giant software
system. It is one clean place for the job, the card, one good
example, and the proof log. If those four pieces are present,
the next run starts faster.
1
The chat or folder
Name it after the job, not the tool. Example: Video Idea Prep, not ChatGPT stuff.
2
The workspace card
Keep the one job, inputs, output shape, rules, and proof step at the top.
3
One good example
Save one answer that sounded right so the workspace has a target to match.
4
The proof log
Write what changed, what was guessed, and what you checked before using it.
Help me build this AI workspace like a small workbench.
Workspace card:
[paste the card]
Good example I like:
[paste one answer if I have one]
Return:
1. the clean workspace name
2. the exact rules to pin at the top
3. the first prompt I should use inside it
4. the proof log I should keep after each run
5. the one thing I should test on the second run
Keep it beginner-friendly and do not turn this into a giant system.
Step 6
Filled example: Video Idea Prep workspace.
This is the kind of workspace Chance can use for YouTube without
sounding fake. The goal is not to let AI make the whole video.
The goal is to turn messy research into a clearer hook, talking
points, and one thing to explain better.
Workspace name
Video Idea Prep
Use this when a rough idea, transcript note, or saved research file needs to become a plain-spoken video outline.
Inputs
Topic, audience, what Chance already knows, links or notes to verify, and what he wants people to understand.
Output
One hook, three talking points, one simple example, one phrase to avoid, and one proof check.
Voice rules
Plain, direct, curious, no fake guru tone, no made-up numbers, and no pretending to understand a source that was not provided.
# Filled Workspace Card - Video Idea Prep
One job:
Turn messy AI research or YouTube notes into a video outline Chance can actually talk through.
Inputs I bring:
- rough topic
- notes or transcript chunks
- what I think it means
- who I am trying to help
- anything that needs verification
Output I want:
1. simple hook
2. three plain talking points
3. one real-life example or analogy
4. what I should not claim yet
5. one proof check before recording
Rules:
- do not make me sound smarter than the facts support
- do not invent what a video, tool, person, or company said
- explain acronyms in normal language
- keep the words close to how I would say them
- mark what I should verify before I teach it
Proof step:
Tell me what came from my notes, what you inferred, and what I still need to confirm.
Step 7
Run it twice before you call it a real workspace.
A workspace is not proven because it worked once. Run a similar
second task through it. If the second run is faster and safer,
the workspace is worth keeping. If not, tighten the job.
Run 1
Get the first useful draft.
Use a real task with private details removed. Save the answer only if it actually helps.
Fix
Tighten the rules.
Add the missing voice rule, proof check, example, or do-not-invent line you needed.
Run 2
Try a similar task.
If it saves setup time on the second run, keep the workspace and use it again next week.
Review this workspace after the second run.
Workspace name:
[name]
Run 1 result:
[what worked / what failed]
Run 2 result:
[what worked / what failed]
Tell me:
1. what rule I should add
2. what example I should save
3. what proof check is missing
4. whether this workspace is worth keeping
5. how to make the job smaller if it is still messy
Step 8
Save the steps, not just the prompt.
A saved prompt is good. A saved workflow is better. Write down
what you paste, what you ask for, what you check, and what you
do with the answer.
Input
What messy material do you give it?
Prompt
What do you ask it to make?
Follow-up
What short fix usually makes it usable?
Check
What facts or promises must you verify?
Workspace checklist:
1. Paste the non-private notes.
2. Ask for the output shape I need.
3. Ask for a shorter or clearer version.
4. Check facts, dates, prices, names, and promises.
5. Save the version that worked.
6. Write down what I changed before using it.
Step 9
Let the helper work inside boundaries.
AI can help more when it has context, but context does not mean
dumping everything into the chat. Give enough to do the job and
keep anything sensitive out unless you have a real reason.
Safe boundaries
Use summaries instead of full private records.
Remove customer names when the name does not matter.
Keep passwords, payment details, and account access out.
Ask for drafts, checklists, and questions first.
Slow down when...
The answer will go to a customer.
The work involves money, refunds, guarantees, or contracts.
The topic is legal, medical, tax, or compliance related.
The tool suggests publishing or sending without review.
Step 10
Require proof before trust.
Do not just ask, "Is this good?" Ask the tool to show what it
used, what it changed, what it guessed, and what a human should
check. That is how beginners stay in control.
Green light
The answer is useful, plain, checked, and still leaves the final call with you.
Red light
The answer invents details, skips verification, or sounds ready to send before you approve it.
Step 11
Run one small rhythm each week.
A workspace becomes valuable when it gets used again. Keep the
rhythm simple enough that a beginner can do it without feeling like
they joined a complicated software class.
Bring
Bring one real job, not a vague idea.
Build
Use the card and rules sheet to get the draft.
Prove
Ask what changed, what was guessed, and what needs review.
Post
Share the card, result, and one question in the Classroom.
Save
Keep the better version so next week starts faster.
Weekly workspace check-in
Workspace I used:
Real job I brought:
What AI helped with:
What I verified:
What I changed before using it:
One rule/example I saved for next time:
My question for the Classroom:
Plain rule
Keep one job you can check. If the workspace does not save time on the second use, make the job smaller and tighten the rules.
Step 12
Use this as the bridge into the Classroom.
The Field Guide gets the first useful answer. The AI Workspace
chapter turns that answer into a repeatable habit. The Classroom
is where that habit can get weekly structure once Chance opens
the join path.