用 Grok Bot 做外呼该建几个代理

Matt @mattceras原推发布于 2026.10.22

在 X 上打开原推

Been talking to a lot of ppl about using grok bot for outbound and the most common question from new users is How many agents should you have? Why can't you just use one main chief of staff agent and what should each of them do? The best way I've found to do this is to split them up based on tasks and think of them like a small team of people If you have a pod at your agency each agent is basically a person in that pod with their own role The main ones I would create: - Campaign monitor, does domain checks against blacklists, inbox health and reply rate by domain so you can actually see whats going on. Also checks when campaigns are about to run out of leads - Copywriter, knows how to write copy. Train it on whatever frameworks you use or just your historical data from your sequencer - Reply agent, for some reason I've noticed it works best if you still webhook from the sequencer into a slack channel and have a cursor cloud agent in that channel that wakes grok bot up Then manually go through the first 50-100 replies with it so it learns how you respond, how you suggest times to meet, how you handle certain objections etc. Train it like you would a person - Intent/research agent, scrapes public directories, reddit, google searches, whatever info you need from the world to write about a certain offer The reason you want a bunch of these is each one has deep context on one specific task The reason you don't want to run it all through one agent is it can only do so many things at one time If you have an agent doing research for a campaign it can't also be replying to leads and managing the inbox At any given time you might want to talk to your main chief of staff agent about whatever. Replies are coming in and need to go to the reply agent. Then you have a separate agent researching a client offer you're getting ready to launch campaigns for All of that has to happen at the same time so splitting them up by task is very helpful