The AI agent wasn't useful until I defined where it had to stop
retained text
I'm Marina Pasqual, an applied AI researcher and co-founder of VELIN PRIVÉ, a personal and business matchmaking agency. We use AI to reduce operational routine, but not to make the human decisions that define matchmaking. I applied the same principle to an agent I built for community growth. It helps me find and qualify potential community members on LinkedIn and keeps the outreach pipeline organised in Google Sheets. It can research candidates and prepare the next step, but post-connect messages still wait for my approval because relevance and tone need human judgment. That boundary mattered more than making the agent more autonomous. Once the stopping point was explicit, it became much easier to see whether the system was reducing work or only creating more review. For people running agents in real operations: what does yours do independently, and what always waits for a human?
document fetched 2026-09-26 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1wpryq1/the_ai_agent_wasnt_useful_until_i_defined_where