Lesson from a couple of months running Efficiency Sprint on Claude Code agents: every agent runs until it hits a gate a person still owns. Mine is "ok to send". Agents draft all day. Nothing leaves until I type it. Where's your gate?
@Michael_Fenech_ Number 1 pays for the rest. In the companies where I ran ops, quotes nobody followed up on were the biggest leak. The prompt is easy. Someone still has to own the follow-up once the draft is approved.
I'm looking for owners who tried AI in their company and got burned. A pilot that went nowhere. An agent that made things up. What happened? Tell me in the replies.
@jasonlk Same curve on the ops side. Building the agents stopped being the hard part a while ago. What still eats my week is deciding who approves what and what happens when two of them disagree.
Check out my latest article: Is your company ready for AI? Your tools most likely are. Your operations probably aren't https://t.co/MawxSbXliv via @LinkedIn