You will run the full process on a real meeting and compare the time with your usual note-taking.
Jun Wei runs a monthly review call with a client. Writing the follow-up usually takes her 40 minutes from the end of the call to sending it (example figures). This time she ran the whole AI meeting-notes process on that call and timed it, and you're going to do the same with one of your own meetings.
The earlier lessons in this module each covered one part of the job. Lesson 7.1 covered permission and an approved tool, Lesson 7.2 a summary prompt, and Lesson 7.3 checking. Now you run all three on a single real meeting, from the invitation to the follow-up. Then you compare the time with your baseline: how long your meeting notes or follow-up usually take without AI, from the end of the meeting to sending.
Pick a meeting in the next week or two where you would normally write the notes or the follow-up yourself. An internal project meeting is a good first choice. A client meeting is acceptable only if your company's tools and rules allow it and the client agrees.
Go back to the notes you made in Lesson 7.1. Confirm which transcription tool you'll use and that the company approves it for this kind of meeting. Mention the recording in the meeting invitation. At the start of the meeting, before anything substantive is said, give a one-sentence notice that you are recording.
If anyone objects, don't record. Take notes the usual way, log the meeting as a non-AI entry, and pick another meeting for the exercise. A refusal is a normal outcome, and handling it smoothly is part of doing this properly.
During the meeting, take light notes on decisions, actions and anything that felt unresolved. You'll use them later to check what the transcript missed.
Start the timer when the meeting ends. Run the summary prompt from Lesson 7.2 on the transcript, then check the result with the routine from Lesson 7.3. Check every decision and action item against its timestamp. Ask whether it was agreed or only suggested, whether the owner is right, and whether the deadline was really stated. Compare the summary with your own notes. Then do a separate pass on names, numbers and dates, and confirm them against the invitation, documents or calendar where you can. For a client meeting, remove anything said off the record. Where the summary has a gap, mark it rather than guess.
Record every change in a corrections list. For each one, write what the summary said, what it should have said, and which kind of error it was. Each error falls into one of these types:
suggestion turned into decision wrong owner invented deadline dropped disagreement transcription error something missing
Send the checked summary to the participants and stop the timer when it's sent. Your total time runs from the end of the meeting to sending the checked summary. That total is the figure you compare with your baseline.
Over the next day or two, note any corrections participants send in reply. Add them to the end of your list, marked as found by others. Say a colleague replies "actually I said end of month, not end of week". They have found an error your check missed, and you should know about it.
Then log the entry: total time against your baseline for meeting notes, with the AI column marked yes.
Running the prompt took her 2 minutes, and the result arrived 1 minute later. She spent 12 minutes checking against the timestamps and her own notes, 6 minutes on fixes and 2 minutes on a final read, then sent it. That came to 2 + 1 + 12 + 6 + 2 = 23 minutes against her baseline of 40, a saving of 17 minutes (40 - 23 = 17).
Her corrections list had four items. The assistant had turned "early next month" into a specific date, so she changed it back to what was said (invented deadline). The client company's name had been transcribed as two separate words (transcription error). A disagreement about the number of participants per session had been left out, so she added it as an open question (dropped disagreement). A client remark about their budget approval process had been made off the record, so she removed it.
The next day, the client's HR manager replied with one correction: the venue contact was a different colleague from the one named. She added it to the list, marked as found by the client.
Five errors in one summary can look discouraging. Jun Wei doesn't see it that way. Every one of them would have been sent if the summary had gone out unchecked, and she caught four of the five in 12 minutes of checking.
If your first attempt is slower than your baseline, keep the result anyway. Checking gets faster with practice, and Module 9 looks at how to read early results like this.
When you finish, you should have three things. The first is a checked summary, already sent, with decisions, owners and deadlines, and with gaps marked rather than guessed. The second is a corrections list showing what changed and why, with participant corrections added at the end. The third is a log entry with your total time against the baseline for meeting notes, with the AI column marked yes.
Pick your meeting now and add the recording notice to the invitation.
Deliver a checked meeting summary with decisions, owners and deadlines, a list of corrections you made, and a time comparison with your baseline.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).