TPM 判断力 vs AI 数据¶
状态: ✅ 已完成
创建日期: 2026-03-27 最后更新: 2026-03-28
LinkedIn Post¶
AI now does my meeting prep. 3 hours down to 4 minutes. The meeting is still just as long.
Faster data didn't make the calls easier. It got me thinking about what I picked up over the years — what AI can take over, and what it can't.
Singles' Day, years ago. 430 engineers. $30.8B. 1,359 contingency plans loaded. Dashboards all green.
All green was the problem.
I pulled one lead aside and asked: "What's not in this report that keeps you up at night?" Forty minutes later I had the only risk that mattered that week. No dashboard caught it. I had to ask.
That kind of work, AI still can't do.
Six years of war rooms taught me this: someone tells you "we're fine" at 11pm, but their face says otherwise. Which question they answer fast, which one they stall on — I learned to read that. Haven't seen a model pick it up.
Another time, payments went down. Two teams blaming each other. Even with today's AI, the dependency map could be perfect and it still wouldn't tell you: give them 20 more minutes, or escalate now? Get it wrong and it's on you.
These days I use AI to pull data ten times faster than before. But the call is still mine. The judgment I built in those rooms — that's the part that's hard to replace.
If you're in Vancouver or Seattle: what's the last call you made that no AI could've made for you?
TPM #AgenticAI¶
X Standalone Post¶
AI cut my prep time. The judgment calls didn't go anywhere.
In Vancouver or Seattle: what's a war room call no AI could've made for you?
TPM #AgenticAI¶
中文翻译¶
现在 AI 帮我准备周会，从 3 小时缩到 4 分钟。但会还是那么长。
数据更快了，判断没变轻松。这让我一直在想以前那些年攒下来的东西，到底哪些是 AI 能接的，哪些不是。
双十一那年，430 个工程师，308 亿美金。1,359 个应急方案挂着，大屏全绿。
全绿就是问题所在。
我找了一个负责人，就问了一句："报告里没写的东西，你最怕什么？" 聊了 40 分钟，捞出来那周唯一要命的风险。看板上根本看不到。得自己去问。
这种活儿，到今天 AI 也接不了。
六年作战室教会我的事：晚上 11 点有人跟你说"没事"，脸上写着"有事"。他哪个问题秒答、哪个问题卡壳，我能看出来。没见哪个模型读得出这个。
还有一次支付挂了，两边互相甩锅。就算当时有今天的 AI，依赖关系图画得再清楚，该不该再给他们 20 分钟还是现在就喊人 — 这个它替不了你拍。拍错了你兜着。
现在我用 AI 拿数据，比以前快十倍。但拍板的还是我。那些年攒的判断力，才是真正没法被替代的东西。
在温哥华或西雅图的朋友：你最近做的哪个判断，是 AI 替你做不了的？
发布记录¶
- LinkedIn: https://www.linkedin.com/feed/update/urn
share:7443332174352936960/ （2026-03-27 发布） - X: https://x.com/XINDR365/status/2037567448686481411 （standalone + Reply 放 LinkedIn URL）
- 配图:
10-images/linkedin_post_tpm_judgment_vs_ai_20260327.png（红色剪影概念图）