Microsoft Copilot can now directly help you create:
Word.
Excel.
PowerPoint.
Real files.
But don’t start your prompt with just:
“Make me a presentation.”
Today, spend just:
One minute.
Writing these four fixed lines at the beginning of your prompt:
Output.
Audience.
Must keep.
No fill-in.
The purpose of these four lines is not to:
Write a fancy prompt.
But rather to:
Clarify the work boundaries upfront.
First Line: Output
Start by specifying:
Exactly what you want to create.
Not just saying:
“Organize this.”
But rather:
“Create a 6-page PowerPoint.”
“Create a 2-page Word Project Brief.”
“Create an Excel workbook with a monthly summary.”
Why say this first?
Because with the same data,
Different final uses like:
Manager presentations.
Client proposals.
Internal records.
Budgets.
Require completely different structures.
So the first line answers:
What is the final deliverable?
Second Line: Audience
Next, state:
Who this output is for.
For example:
Managers.
Clients.
New employees.
Finance staff.
Project team.
This line is short but very important.
Because:
A presentation for managers
Is not the same as
A document for colleagues who execute the work.
Managers likely need:
Conclusions.
Risks.
Next steps.
Colleagues might need:
Details.
Owners.
Dates.
Procedures.
So the second line answers:
Who will use or view this output?
Third Line: Must Keep
This step is often overlooked.
List out:
Content that AI should not remove or shrink.
For example, must keep:
The three customer-confirmed requirements.
The official event date.
The current budget.
Known risks.
Next steps.
If it’s Excel, must keep:
Original income.
Expenses.
Months.
Categories.
Total sums.
If it’s PowerPoint, must have:
Problems.
Data.
Recommendations.
Risks.
Next steps.
In other words:
Which items must not disappear even if the layout looks cluttered?
Fourth Line: No Fill-In
Finally, add one sentence:
“Do not add any dates, amounts, numbers, commitments, or conclusions that are not in the data.”
This habit is highly recommended.
Because when AI produces documents,
The easiest trap to fall into is:
It organizes everything so neatly with:
Headings.
Paragraphs.
Tables.
Nice-looking presentations.
But:
Completeness
Does not mean:
The data really exist.
For example, if you only have three customer files
Containing:
Event date.
Number of participants.
Venue.
But not:
Official budget.
If you just say:
“Make me a complete event proposal.”
AI may try to fill in to make it look complete:
Budget structure.
Possible items.
Estimates.
If this is a brainstorming document,
That’s fine.
But if it’s meant
To be sent to a client,
That’s dangerous.
So always say first:
Don’t fill in amounts without sources.
The four lines combined might look like this
Output:
Create a 6-page PowerPoint.
Audience:
For managers attending tomorrow afternoon's meeting.
Must keep:
The customer’s three main requirements, current timeline, two risks, and next steps.
No fill-in:
Do not add any dates, amounts, numbers, or customer commitments not in the data.
Then paste the:
Actual:
Meeting notes.
Data.
Files.
That’s enough.
No need to write a 500-word prompt upfront
Microsoft officially recommends:
Describing:
What file to create.
What content is needed.
Length.
Tone.
If the Agent needs more:
It will ask clarifying questions.
So this 4-line framework
Is not:
A replacement for all prompt techniques.
It just first ensures:
The four most common causes for rework are clearly defined.
Why does specifying “Audience” reduce rework?
Say you ask:
“Make me an 8-page AI introduction presentation.”
Without specifying:
Who it’s for.
AI might create:
What AI is.
Generative AI.
Market trends.
Tools introduction.
Looks fine.
But if the audience is actually:
The boss of a 5-person company,
What they really want to know is:
What step to take first?
How long will it take?
What are the risks?
Who approves?
Is it worthwhile?
The content direction is completely different.
It’s not that AI can’t do it,
It’s that:
You didn’t tell it who the audience is.
Why is “Must Keep” more useful than “Please organize completely”?
When AI summarizes or makes presentations,
It will inevitably:
Make trade-offs.
Ten pages of data
Can’t all fit into
A 6-page presentation.
So deciding upfront:
What can be condensed.
What cannot be condensed.
For example, if the client says:
“Must be completed before October 15.”
This
Cannot be shortened to
“Complete by October.”
Because:
The precision changes.
So:
Essential dates.
Official numbers.
Constraints.
Conditions.
Commitments
Are best listed
Under “Must keep.”
“No Fill-In” doesn’t mean AI can’t reason at all
Clarify this carefully.
You can still request:
Analysis.
Organization.
Comparison.
Recommendations.
For example:
“Based on these three risks, propose three possible solutions.”
This is an:
Analytical task.
No problem.
But if the data lacks:
“The client agreed to delay by one day.”
AI must not turn:
Recommendations
Into:
Confirmed facts.
The real limitation is:
Don’t present guesses as known information.
Word is especially well-suited for these four lines
For example, when making a:
Project Brief, you can write first:
Output:
2-page Word Project Brief.
Audience:
For new project members to read.
Must keep:
Project goals, scope, deadline, owner, current risks.
No fill-in:
Don’t add unconfirmed dates, owners, or budgets.
Then give the original data to the Word Agent.
AI will produce:
A first draft.
You can then give Word:
Final edits.
Excel’s “No Fill-In” is even more critical
Because Excel:
Looks most trustworthy.
Tables.
Formulas.
Charts.
Percentages.
Once created,
It feels like
“It must be calculated.”
So if you use Excel Agent,
The fourth line can be stricter:
“Leave blank any field missing original data; do not substitute estimated values.”
For example:
If September costs are missing,
Leave them blank.
Don’t guess values just to make the chart
Look complete.
PowerPoint most requires “Must Keep”
Because presentations
Always involve significant condensation.
So you can write:
Must keep:
Official figures.
The basis for conclusions.
Three main risks.
Source limitations.
Next steps.
This helps AI understand when compressing that:
These items
Cannot be removed just to make slides look nicer.
After the Agent finishes, first check the Preview
Microsoft’s official workflow includes:
Review Preview.
Do not skip this step.
Don’t just say:
“Wow, it really made the PowerPoint.”
First ask yourself four questions.
1
Output:
Is it in the format you specified?
2
Audience:
Is the tone and information depth appropriate?
3
Must keep:
Is everything you required still there?
4
No fill-in:
Did AI add anything not present in your original data?
This becomes a four-point QA checklist
That perfectly mirrors your four-line prompt.
The same rules control both Input and Output
Before generating:
Define with four lines.
After generating:
Check with four lines.
This is better than just saying:
“Please review.”
Because “review” leaves AI guessing:
What to check?
Grammar?
Formatting?
Content?
Numbers?
Completeness?
With these four fixed points,
You get clear criteria.
Don’t redo everything if the first draft isn’t perfect
For example:
The PowerPoint structure is roughly correct,
But:
Page 3 is too long.
Page 5 misses risks.
Don’t redo the whole prompt.
Microsoft also supports:
Giving high-level feedback
On an already generated draft.
For example:
“Make page 3 half as long.”
“Add one slide for risks.”
“Move the conclusion to page 2.”
Then return to PowerPoint
For final detailed editing.
The real value of AI is:
Not:
Getting the final version in one shot.
But:
Making the first draft closer to what you need than starting from blank.
Don’t think AI is dumb if it asks questions
For example, if it asks:
“Who is this for?”
“How many pages are needed?”
“Formal or informal tone?”
This actually means:
Your original request lacks needed context.
Answering these
Usually yields better versions than just telling AI:
“You decide.”
These four lines aren’t just for Microsoft Copilot
Any AI that creates:
Reports.
Presentations.
Spreadsheets.
Proposals.
Summaries.
Documents
Can use this framework.
Because it solves the issue of:
Work specifications.
Not:
Microsoft-specific syntax.
You can even:
Save it as your own fixed template.
The fixed template has just four fields
Output:
What do I want to produce ultimately?
Audience:
Who will view or use this?
Must keep:
Which content absolutely cannot be condensed away?
No fill-in:
Which info without evidence must be left blank?
Fill these out in 30 seconds,
Then hand data to AI.
This is actually a small workflow improvement
Before:
Let AI do it.
↓
Find errors.
↓
Redo.
↓
Add data.
↓
Edit again.
Now:
Spend one minute defining specs.
↓
AI makes the first draft.
↓
Check four points.
↓
Final polishing in Office.
The real comparison is not:
Spending one more minute on prompts,
But:
How much rework is saved downstream.
You can test this with 5 times in a row
For the next five outputs:
Word.
Excel.
PowerPoint.
Use the same four-line framework.
Remember these two numbers:
First
How much of the first draft can you keep?
Second
How long does it take to finalize after first draft?
If originally:
You needed 25 minutes to fix after AI output,
And after using four-line specs,
You reduce it to 12 minutes,
That means this prompt method:
Really improves workflow.
Not just because it looks more professional.
If total time doesn’t drop, don’t blindly rely on the template
Maybe your work:
Isn’t suitable.
For example:
Highly creative.
Data varies greatly each time.
Formats constantly change.
Then this four-line method
Helps less.
AI workflow’s most important part is:
Testing.
Not:
Believing one prompt fits all.
What you really need to learn today is this one line
Next time you ask Copilot to:
Create Word.
Excel.
PowerPoint,
Don’t just say:
“Make me one.”
First add four lines:
Output: What I want.
Audience: Who will see it.
Must keep: What can’t be missing.
No fill-in: Don’t guess what’s missing in data.
After AI finishes,
Check using the same four lines in reverse.
That one minute,
The real purpose isn’t:
To make AI write better.
But to:
Make the first draft much closer to a workable version.
If you want to know which step of your work can best be handed over to AI first, comment “workflow”.
Today, improve a bit together with AI.
Learn one AI tip every day.
Save a little time every day.
Grow a little stronger every day.
SasaDaily, growing with you.
Recommended reading
AI Quick Q&A｜Can AI-generated content be used directly? Know These 3 Things First
