AI Meeting Minutes Assistant Guide: Speech Transcription, Smart Summarization & Action Tracking

Why Use AI for Meeting Minutes

Do you spend 1-2 hours after every meeting organizing minutes? Listening to recordings over and over, hunting for key points, adjusting formatting — this repetitive work can all be handed off to AI.

The core workflow for AI-powered meeting minutes is just four steps: Record → Transcribe → Organize → Export. The entire process takes about 10 minutes to produce well-structured meeting minutes.

Step 1: Record the Meeting

Recording is the starting point of the entire workflow. Good audio quality directly determines transcription accuracy.

Phone Recording (Best for: small meetings, 1-on-1 discussions)

• iPhone users: Open the built-in Voice Memos app and tap the red button to start
• Android users: Most phones have a built-in recorder app — search for "Recorder"
• Placement tips: Place your phone in the center of the table, 1-2 meters from speakers

Computer Recording (Best for: online meetings, remote collaboration)

• Tencent Meeting / Feishu / DingTalk: Use the built-in recording feature, which automatically saves audio files
• Zoom: Click the "Record" button at the bottom — files save automatically after the meeting
• Windows: Press Win+G to open Xbox Game Bar for system audio recording

Pro tip: Test-record for 10 seconds before starting, and play back to confirm clear audio without background noise. For online meetings, always enable "Record System Audio" rather than microphone to avoid capturing ambient noise.

Step 2: Speech to Text

After recording, you need to convert the audio to text. This is the most critical step in the entire workflow.

Option 1: iFlytek (Recommended: Best for Chinese)

1. Open tingji.com, register and log in
2. Click "Start Transcription" and upload your audio file
3. Select language (Mandarin/English/Cantonese, etc.) and click "Submit"
4. Wait 2-5 minutes (depending on file size) to get your transcription
5. Edit online and export as Word/TXT format

Advantages: 98%+ accuracy for Chinese, speaker diarization support, 1 hour of free transcription.

Option 2: OpenAI Whisper (Recommended: Free, privacy-first)

1. Install Python environment (if you don't have one already)
2. Run pip install openai-whisper
3. Execute whisper your_recording.mp3 --language zh --model medium
4. Wait for processing to complete — auto-generates .txt and .srt files

Advantages: Completely free, runs locally with no data upload, supports 99 languages.

Option 3: Feishu Minutes (Recommended: For Feishu users)

If your team uses Feishu, simply enable the "Minutes" feature during Feishu meetings. Transcription and summary are automatically generated after the meeting ends — no extra steps needed.

Step 3: Use AI to Organize the Summary

Once you have the transcription, paste it into an AI chat tool and use the right prompts to have AI organize it into structured minutes.

Basic Prompt: Quick Minutes Generation

Based on the following meeting transcript, generate structured minutes including:
① Meeting info (time, attendees, topic)
② Discussion points (categorized by agenda)
③ Decisions and action items (with owner and deadline)
④ Next meeting arrangement

Advanced Prompt: Extract Key Decisions

Extract all decisions from this meeting transcript in a table format:
Columns: No. | Decision | Decision Maker | Impact | Action Required
Mark controversial topics with ⚠️

Pro Prompt: Action Tracker

Generate an "Action Tracker" with task, owner, priority (High/Med/Low), deadline, and status — output as a Markdown table

How to use:

1. Open ChatGPT, Claude, or another AI chat tool
2. Paste the prompt above, then paste your transcription text
3. Wait for AI to generate results, and fine-tune as needed
4. Copy the results into your document

Note: When the transcription is long (over 5,000 characters), consider splitting it into sections, or use Claude which supports long context. ChatGPT free tier has length limits — you may need multiple conversations.

Step 4: Export and Share

The AI-generated minutes are usually already structured text. The final step is formatting and sharing with your team.

Quick Formatting Methods

• Copy directly to Feishu/DingTalk docs: Markdown formatting renders automatically
• Export as Word: Ask AI to output Word-compatible format, then paste into WPS/Word
• Export as PDF: Open Markdown in Typora or VS Code, then export as PDF

Team Sharing Tips

• Send to the project group chat and @mention responsible parties to confirm action items
• Sync action tasks to project management tools (Feishu Projects, Notion, Jira)
• Highlight important decisions separately for easy future reference

Complete Workflow Summary

Let's review the entire workflow with a diagram:

Record Meeting → Speech to Text → AI Summary → Export Minutes

Each step has corresponding tools:

• Recording: Phone built-in recorder / Tencent Meeting recording
• Transcription: iFlytek (best for Chinese) / Whisper (free & open source)
• Organization: ChatGPT / Claude / Feishu Minutes
• Export: WPS / Typora / Feishu Docs

FAQ

Q: What if the recording quality is poor?

A: Choose a quiet meeting room, use an external microphone (clip-on mics work best), and test-record for 10 seconds to confirm quality before starting.

Q: How do I distinguish speakers in group meetings?

A: Both iFlytek and Feishu Minutes support "speaker diarization" — automatically distinguishing different speakers. Whisper currently doesn't support this feature.

Q: Meeting content is confidential. Can I still use AI?

A: Use locally deployed Whisper for transcription (data never leaves your computer). For the organization step, use a locally deployed LLM, or desensitize sensitive information before sending to cloud AI.

Q: Are there tools that complete the entire workflow in one step?

A: Feishu Minutes and DingTalk Quick Minutes come closest to "one-click" — automatically generating minutes after meetings end. If you're not in these ecosystems, try Otter.ai (English) or iFlytek (Chinese).

Efficiency Comparison

Manual vs. AI-assisted:

• Manual: 1-hour meeting = 1.5-2 hours of organizing time
• AI-assisted: 1-hour meeting = 10-15 minutes (including transcription wait time)
• Efficiency gain: approximately 6-8x

If you have 5 meetings per week, you can save about 30 hours per month on organizing time. Using that time for more valuable work is the real meaning of AI tools.