Use ChatGPT to Turn a Meeting Calendar into a Review Checklist

A full calendar can make every meeting look permanent. A better first step is a review table that shows what each meeting is for, what decision it supports, and what still needs a human answer.

This is a documentation-based, illustrative ChatGPT workflow. The calendar is fictional. No ChatGPT run or scheduling result is claimed.

What you will make

You will turn a small CSV into a review checklist with these columns:

Meeting
Stated purpose
Decision or output expected
Evidence from the row
Keep, change, or question
Human follow-up

The result is a discussion aid. It does not cancel meetings or decide that a colleague is unnecessary.

What you need

ChatGPT with file uploads available on your account. OpenAI documents CSV and spreadsheet analysis as supported uses, subject to plan and workspace settings. See the Data analysis with ChatGPT help page.
A low-risk calendar export with names and private details removed.
One review rule: a meeting should have a stated purpose and a useful decision, output, or coordination need.

Create meeting-review.csv with this complete fictional input:

meeting_id,meeting_name,cadence,attendees,stated_purpose,expected_output
M-101,Monday sales sync,Weekly,Sales lead; account owners,"Review pipeline movement and blockers","Updated owner and next step for blocked deals"
M-102,Status readout,Weekly,Operations lead; department heads,"Share updates from each department","No output stated"
M-103,Launch decision review,Once,Product lead; marketing lead; support lead,"Decide whether the help centre update is ready","Go or hold decision with named follow-up"
M-104,Project check-in,Weekly,Project team,"Talk about progress","No output stated"
M-105,Customer escalation review,Weekly,Support lead; product lead,"Review unresolved customer issues","Priority, owner, and response date for each open issue"
M-106,Leadership notes,Monthly,Department heads,"Read the monthly notes together","Questions for the next leadership meeting"

These rows are practice material. They are not evidence that any real meeting is wasteful.

1. Upload the file and set the boundary

Open a new ChatGPT conversation, attach meeting-review.csv, and paste the prompt below. Keep the meeting names and attendee labels fictional while learning.

Review meeting-review.csv as a small calendar sample.

Create a table with these columns:
- Meeting ID
- Meeting name
- Evidence from the source row
- Stated purpose
- Expected output
- Keep, change, or question
- Human follow-up

Use these rules:
- Use only the values in the CSV.
- Quote the stated purpose and expected output exactly when you use them as evidence.
- Mark a row as question when the purpose or output is missing or too vague to support a scheduling decision.
- Mark keep when the row states a concrete decision, owner, review, or coordination output.
- Mark change when the meeting may be useful but its cadence, attendees, or format needs a human review. Do not invent a new cadence.
- Never claim that a meeting is unnecessary.
- Never infer the importance of a person from the attendee list.
- Do not calculate time saved or productivity gains.
- Keep missing details as NOT STATED.
- Add a final section called Review checks.

The fixed rules matter more than the label. They keep a vague row from becoming an overconfident cancellation recommendation.

2. Compare the table with the source

For the fictional input, an illustrative review might look like this:

Meeting	Evidence	Decision	Human follow-up
Monday sales sync	Pipeline blockers and a next step are stated	Keep	Confirm the owner field is used each week
Status readout	No output is stated	Question	Ask whether the update can be written instead
Launch decision review	A go or hold decision is stated	Keep	Confirm the decision owner
Project check-in	Progress is stated, but no output is stated	Question	Define the decision or artifact before changing cadence
Customer escalation review	Priority, owner, and response date are stated	Keep	Confirm the escalation criteria
Leadership notes	Questions for a later meeting are stated	Change	Decide whether the notes need a live discussion

This is an editor-written illustrative artifact, not a model result. Compare every row with the uploaded CSV.

3. Run the acceptance checks

The workflow passes when:

all six meeting IDs appear in the output
every quoted phrase matches the source row
M-102 and M-104 remain questions because their outputs are not stated
no row says a meeting should be cancelled
no attendee is judged by their title or presence
time savings and productivity claims are absent
the final table gives a human a specific follow-up question

Troubleshooting

ChatGPT invents a meeting objective. Ask it to replace the invented text with NOT STATED and cite the meeting ID.

It recommends cancellations. Repeat: “Use only Keep, Change, or Question. Never recommend cancellation.”

It drops a row. Ask for one output row per source meeting ID, then compare the six IDs manually.

It rewrites the evidence. Ask for exact source phrases in a separate column. Treat the rewritten sentence as commentary, not evidence.

Limits and decision

A calendar row rarely explains the full value of a meeting. Check team context, decisions made outside the meeting, accessibility needs, and legal or customer commitments before changing a schedule. Use approved or fictional data and remove unnecessary personal details.

TRY: a small, low-risk calendar export with one review owner.

SKIP: automatic cancellations, sensitive personnel discussions, and any schedule change without the people involved.

USE: the table as a meeting redesign agenda, not as a verdict about colleagues.

For one practical AI-for-work lesson each week, join The Weekly Verdict.