註冊並分享邀請連結,可獲得影片播放與邀請獎勵。

Michael | AI + Agents
@Michael_Fenech_
🤖 Building AI agents that do real work. 📈 3 startups, $8M raised, still shipping. 👀 Sharing everything I learn from building.
加入 August 2024
134 正在關注    59.4K 粉絲
When your Grok Bot gets a fact wrong, the most important thing is what it does after you correct it. My Grok Bot told me a meeting was Monday. I corrected it: “It’s Tuesday.” Instead of checking my calendar, it tried to work out the new date itself. And got it wrong again. That exposed a reliability rule I’m now using across my agents: When corrected on a fact that has a source of truth, re-check the source before answering again. Calendar fact → check Calendar. Customer fact → check CRM. Email fact → check Inbox. Order fact → check the database. This is a simple rule worth adding to any agent that works with live business data: Correction → source check → answer. A correction shouldn’t trigger another guess. It should trigger verification.
顯示更多
0
31
130
5
轉發到社區