AI meeting transcription and minutes nanny-level tutorial: from recording to action item closed loop

🛒 A complete guide from registration tools to distribution tracking for meeting-heavy managers, HR, administration and remote teams.

Tutorial objectives and applicable readers

This tutorial teaches you step by step how to set up an "AI meeting records and minutes" process: selecting a transcription tool, recording and transcribing a meeting, using AI to generate minutes, manually checking and distributing them, and tracking action items. The entire process does not rely on programming, and managers, HR, administration, PMO and remote teams with intensive meetings can quickly get started. A complete practical session takes about 30-60 minutes.

1. Preparation Checklist

  • [ ] Register a meeting transcription tool (Chinese is recommended first: Tongyi Listening; or Notta / Fireflies, the tutorial uses Tongyi Listening as the default example)
  • [ ] Prepare a real meeting (either online meeting or local recording file)
  • [ ] Confirm team members’ informed consent to recording/transcription (relating to compliance)
  • [ ] Prepare a minutes distribution channel (Feishu/DingTalk/Email/Notion)
  • [ ] Create a "meeting minutes template" with fields: meeting topic, time, participants, topics, discussion points, conclusions and decisions, action items (person in charge + deadline)

2. Tool registration and settings

  1. Open the Tongyi Tingwu official website and log in with your Alibaba Cloud account (or register according to the official instructions).
  2. Enter the workbench and familiarize yourself with the two entrances of "Recording File Transcription" and "Real-time Transcription".
  3. Confirm that the output language is Chinese in the settings and turn on "speaker separation" (multi-person recognition).
  4. If you need automatic minutes, make sure that the AI ​​minutes/summary function is turned on (the function name is different in different versions, please refer to the interface).

3. Record and transcribe a meeting

Method A: Upload the recording file

  1. Use mobile phone recording or the recording function of the conference platform to record the entire meeting.
  2. Open Tongyi Listening → Upload audio/video files.
  3. Wait for the transcription to be completed (usually within a few minutes, subject to real-time service conditions).

Method B: Real-time transcription

  1. Turn on the browser plug-in or system-level transcription during online meetings (whichever is the access method supported by the selected tool).

  2. After the meeting, view the complete transcript and timeline in the tool.

  3. After the transcription is completed, check whether the "speaker separation" is correct: if multiple people are recognized as the same person, manually split or label them.

4. Use AI to generate meeting minutes

  1. Find the "AI Minutes/Intelligent Summary" entrance on the transcript page (different tools have different names).
  2. After clicking Generate, check whether the output contains:
    • Meeting topic and time
    • List of issues
    • Discussion points
    • Conclusion and decision-making
    • Action items (responsible person + deadline)
  3. If the template fields are incomplete, copy the following prompt words and let AI rewrite them according to the template:
Please output the minutes according to the template based on the following meeting transcript:
Template fields: Meeting topic | Time | Participants | Topics | Discussion points | Conclusions and decisions | Action items (responsible person + deadline)
Requirements: Only rely on the content of the transcript and do not add information that has not been mentioned; list the person responsible and the time for the action items one by one; mark [to be confirmed] if it cannot be determined.
  1. Paste the AI output into the minutes template and save it as a first draft.

5. Manual checking and correction

It is not recommended to distribute AI minutes directly. Please check before publishing:

  1. Fact Check: Check the names, numbers, dates, and decision-making caliber against the transcript.
  2. Action Item Check: Confirm item by item whether the person in charge and the deadline are consistent with the discussion at the meeting. If there is no clear statement, mark it [to be confirmed] and confirm it after the meeting.
  3. Sensitive Information Screening: Delete or desensitize private and confidential content, and confirm the scope of distribution.
  4. Terminology Correction: Correct product names and professional terms that are incorrectly transcribed by the tool.
  5. Save as final draft and mark the reviewer and date.

6. Distribution and action item tracking

  1. Distribute the final minutes to team collaboration tools (Feishu/DingTalk/Email/Notion).
  2. Convert action items into tasks: set a responsible person and deadline for each item.
  3. Set reminders for each action item in the calendar/task tool.
  4. Do a review of action items every Friday: confirm the completion status one by one, and clarify the reasons and new deadlines for unfinished items.

7. Verification method

  • Complete transcription: key content of the entire meeting can be found in the transcript
  • Accurate minutes: Spot check 5 key points for consistency with the recording
  • Action items are executable: each item has a responsible person and deadline, and can be tracked
  • Distributed in place: all relevant parties receive and confirm the minutes
  • Convenient search: you can find the conclusion of a certain meeting within one minute after the meeting

8. Frequently Asked Questions and Troubleshooting (FAQ)

  1. Is the recognition inaccurate and there are many typos? Use a clear audio source; register before speaking when multiple people are speaking; provide a glossary of important terms before the meeting; key information is manually corrected against the recording.
  2. Did the AI minutes make up things that were not said at the meeting? The prompt word is mandatory "Only based on the transcript, no additional information"; cross-check with the transcript before publishing.
  3. The meeting involves privacy, can it be used? Please check the service provider's privacy terms and storage area first; confidential meetings will not be transcribed; only local recordings will be retained if necessary.
  4. Poor multi-language/Chinese and English mixed recognition? Choose a service that supports multiple languages; or process it by segment; retain manual proofreading for important meetings.
  5. The free quota is not enough? Only high-value meetings are transcribed; for ordinary meetings, the platform comes with built-in minutes or only key points are recorded manually.
  6. Action items are always followed by no one? Synchronize action items to task tools and set reminders. Weekly meetings will review the closed loop to form a habit.

9. Advanced expansion

  • Connect to CRM/Project Management: Let the output of sales meetings and project meetings be written directly into the business system to reduce secondary entry
  • Meeting Knowledge Base: Precipitates product decisions, customer feedback, and training content for team retrieval and reuse
  • Insight Analysis: Perform topic clustering on historical meetings to discover high-frequency issues and collaboration bottlenecks
  • Linked Production: Meeting minutes are input into the AI writing and knowledge retrieval solution as material, and the meeting output is directly deposited into content assets

10. Practical example: the entire recording process of a complete meeting

Taking a 45-minute "Product Iteration Weekly Meeting" as an example, let's go through the process:

  1. Before the meeting: Mark "Transcription required" on the meeting calendar; turn on real-time transcription in Tongyi Listening (or agree to use the conference platform to record).
  2. During the meeting: The host will start with a reminder that "this meeting will be transcribed for minutes". Participants must sign up before speaking to ensure the quality of recognition.
  3. After the meeting: The transcription is completed, check whether the speaker separation is correct; correct the paragraphs with many typos against the recording.
  4. Generate minutes: Use the AI ​​minutes function to generate a first draft, and check whether it contains topics, discussion points, conclusions and action items according to the template.
  5. Manual verification: Focus on checking the "Conclusions and Decisions" and "Action Items" columns - for example, "The new version of the details page will be online before August 20, the person in charge = Xiao Ming", and compare it with the transcript one by one.
  6. Distribution and follow-up: Minutes are sent to the project group, action items are converted into tasks and reminders are set; the closed loop is confirmed one by one during the review on Friday.
  7. Precipitation: Store this round of decisions in the product knowledge base for subsequent retrieval.

Self-inspection after completion: if all five key decision-making points can be found in the minutes, and each action item has a responsible person and time, it is deemed to have met the standards. Consolidate this set into team weekly meeting SOP and reuse it every week.

11. Meeting minutes template (can be reused directly)

#meeting minutes
- Meeting topic: {topic}
- Time: {date time}
- Participants: {list}
- Recorder: {Name} / Checker: {Name}

## Topic
1. {Topic 1}
2. {Topic 2}

## Discussion Points
- {Key points, you can quote the speaker}

## Conclusion and decision-making
- {Conclusion 1} (Source: {Who proposed it, whether it was unanimously approved})

## Action items
| Matter | Person in Charge | Deadline | Status |
|---|---|---|---|
| {Item 1} | {Responsible person} | {Date} | To-do |

## To be confirmed
- {Items that cannot be confirmed and need to be verified after the meeting}

Save this template to a team document and apply it to every meeting. Use the transcription tool to achieve "continuous meetings and follow-up".

12. Summary: Make meetings truly produce value

The value of a meeting is not in "what happened", but in "what was left and what was followed up". Use transcription tools to digitize meetings, use AI minutes to extract conclusions and action items, use manual verification to maintain accuracy and compliance, and rely on task tracking to close the loop on action items. Once this process is solidified, the team's meeting output will change from "notes scattered among everyone" to "searchable team assets." Start with a high-frequency meeting first, and then gradually promote it to all meetings worth keeping. You will quickly see a double improvement in meeting efficiency and implementation rate.

User Reviews

  • Loading reviews...