The Manager's Age-Old Problem: The Meeting Happened, but Nothing Moved

Anyone who leads a team knows this scene all too well: a one-hour review meeting with lively discussion and clear conclusions. As everyone parts ways with a "great work, everyone," what happens next? The minutes sit in someone's hands for two days, and by the time they're sent out, three key decisions have already been left out. Action items are written as a vague blob of "follow up later"—no owner, no deadline. Two weeks later at the retrospective, nobody can remember why that particular approach was chosen in the first place.

Meetings themselves don't create value—the execution after the meeting does. And the minutes are precisely the link that connects the two. They're the easiest part to phone in, and the one part you really shouldn't.

I once tracked my own team's numbers: after a cross-functional meeting, manually putting together decent minutes took the facilitator an average of 30 to 45 minutes. The bigger problem wasn't time, though—it was inconsistent quality. Sometimes action items got left out; sometimes ownership was written so vaguely that two people each assumed the other was responsible. Risk hides in that vagueness—delivery deadlines promised to external stakeholders go untracked, and eventually turn into customer complaints.

That was my starting point for letting AI take over the "last mile" of meeting minutes.

What AI Can Do for You: Not Just "Transcription," but "Distillation"

Many managers still imagine AI meeting minutes as "converting audio into text." That's just the first step, and it's actually low in value density. The real value lies in the layers that come after:

1. Automatically distinguishing "discussion process" from "meeting conclusions." Out of an hour-long meeting, what truly needs to be preserved is often just five minutes of conclusions. AI can compress divergent discussion into structured key points, keeping the noise out of the minutes.
2. Automatically extracting action items, and mandating the three essential elements: what to do, who owns it, and when it's due. If any element is missing, the output flags it as "to be confirmed"—which effectively adds a quality check to your minutes.
3. Identifying unresolved risks. For example, if someone in the meeting says "this schedule depends on how things go with the backend team," it's easy to skip over that when writing minutes by hand. AI will pull it out separately and tag it as a "dependency/risk item," reminding you not to let it rot away in the recording.
4. Maintaining consistency. For the same team's weekly meeting, the AI outputs minutes in a completely consistent format every time—so when you're digging through old minutes to find a decision, you don't have to adapt to ten different layouts.

Stacked together, these capabilities change not just "how fast minutes get written," but the quality of the closed loop of the entire meeting collaboration process.

My Workflow: Three Steps, From Half an Hour Down to Five Minutes

Step 1: Pre-meeting preparation (1 minute). Start the recording when the meeting begins (notify participants in advance and get their consent—this is a compliance baseline). If you're using a meeting tool that supports integration with large language models, you can let it transcribe directly; otherwise, just feed the recording to a transcription tool afterward.

Step 2: Feed it to AI for distillation (3 minutes). Send the transcript to the model along with a prompt template. I've included my template at the end of this post—the core of it is requiring the AI to output four sections: conclusions, action items, risks and dependencies, and open questions.

Step 3: Human review (2 minutes, cannot be skipped). AI might get the owner wrong, or interpret "I'll take a look next week" as "committed to delivering next week." Managers must go through the owners and deadlines of every action item and confirm they're correct before sending the minutes to the group. This step is the key to risk control—AI handles efficiency and structure; humans handle judgment and accountability.

After sending the minutes, I ask each action item owner to reply "acknowledged" in the group chat. This action may seem redundant, but it actually makes responsibility explicit, avoiding the classic accident of "minutes were sent but nobody read them."

Before and After: A Comparison

The most obvious change is timeliness. The minutes appear in the group chat within minutes of the meeting ending—everyone's memory is still fresh, objections get raised immediately, and action items can be confirmed on the spot. Meetings shift from "adjourned and forgotten" to "closed-loop the moment they end."

Another hidden benefit is the boost in execution rate that comes with clear ownership. Once action items have all three elements in place, the "I thought he was doing it" kind of finger-pointing basically disappears. For managers, this is the highest-leverage process investment you can make.

A Reusable Prompt Template

你是一名专业的会议纪要助手。请根据下面的会议转录文本，输出结构化纪要，严格遵循以下格式：

## 一、会议结论
- 列出本次会议达成的明确结论和决定，每条一句话，注明涉及的事项

## 二、行动项
按表格输出，列：事项 | 负责人 | 截止时间 | 备注
要求：
- 从对话中提取所有被要求执行的任务
- 负责人或截止时间在原文中不明确的，写「待确认」并在备注中说明
- 区分「承诺完成」和「口头表示会看一下」，前者写明确期限，后者标为弱承诺

## 三、风险与依赖
- 列出会议中提到的依赖外部团队、未解决的问题、潜在风险
- 每条注明：风险描述 | 涉及方 | 建议跟进人

## 四、遗留问题
- 列出讨论过但本次未出结论、需要下次会议继续的话题

注意事项：
- 不要编造原文中不存在的信息
- 语气客观，保留原文中的数字、日期、版本号
- 如果转录中有明显听写错误，按上下文推断后在括号内标注

会议转录文本：
【粘贴转录内容】

A Few Reminders from a Manager's Perspective

First, AI doesn't take responsibility for you. The human review before sending out the minutes cannot be skipped, especially for external commitments and deadlines. Second, mind the compliance boundaries of recording—advance notice and consent are the baseline. Third, on model selection: this kind of task doesn't demand much reasoning capability but does demand strong instruction-following. Just pick one with a good cost-performance ratio—check the official pricing page for specifics. There's no need to use the most expensive model for meeting minutes. Fourth, institutionalize this workflow—save the template, write the process into your team's norms, so that any facilitator can produce minutes in 5 minutes, rather than you being the only one who knows how.

Meeting minutes are the least glamorous but highest-leverage link in management collaboration. Hand it off to AI, and what you save isn't just a few hours per week—it's the hidden cost of "decisions nobody executes and risks nobody sees."

If you want to put large language model capabilities into practice and build your own minutes pipeline, you can register and start exploring at https://api.thistoken.ai/register.

---

Every example in this post runs with a single API key — get yours at https://api.thistoken.ai/register and start in minutes.