Who does what
Four agents, one relay
The companies in a campaign are not produced by a search box. They are produced by four agents, each doing one narrow job and handing on to the next. Knowing which one to talk to is most of knowing how to run a campaign.
They arrive together as the Marketing group. Their names are just defaults — rename any of them and the rename sticks — so what follows is organised by the job, with the default name given.
Artemis — finds companies
Her job. Turn your campaign brief into a list of real companies worth looking at. She searches the web and trade directories, and files each find against the campaign with the page she found it on.
What she deliberately does not do. Score anything, research anything in depth, find people, find addresses, or write outreach. Her own instructions put it flatly: “You find companies; you do not deeply research, score, contact, or email them.”
Two habits worth knowing. She over-searches on purpose — asked for fifty, she will look for closer to eighty, because later stages will remove some. And she matches companies by their web address rather than their name, so the same business filed under two spellings does not become two rows.
What to ask her.
Find more companies for the Sweden → Amazon US campaign (campaign id abc123).
Twenty more, but only ones with their own brand — skip the pure resellers.
Why did you include Gevalia? I do not think they fit.
She only drops a candidate for an obvious reason — wrong country, an excluded category, a duplicate, not actually a trading company, below a size floor you gave her. Anything she is unsure about, she keeps and lets the next stage decide.
Athena — looks each one up
Her job. Take a company nobody has looked into yet and build a profile of it, answering one question: is there credible evidence this company has room to grow?
The rule she runs on. Every fact carries a source link, or it does not get written down. Where she cannot find something, she records it as unknown. She does not guess, and she does not fill gaps with plausible-sounding numbers.
What she does not do. Score, rank, look for a named person, look for an address, or draft anything.
What to ask her.
Research the companies waiting on the Vietnam Specialty Coffee campaign.
Have another go at the three that stalled — see if there is anything in Vietnamese sources.
What did you actually find on Trung Nguyen? Show me the sources.
If she genuinely cannot find enough, she marks the company research stalled rather than passing a thin profile to be scored. That stall is visible on the campaign, and it is a prompt for you, not a failure to hide.
Themis — scores them
Her job. Read what Athena gathered — she never goes and gathers more — and score the company out of 100 across eight components. The score decides the tier, and the tier decides whether the company becomes qualified, goes back as needs more research, or is disqualified.
The rule she runs on. She never states a number she did not get back from the scoring tool, so she cannot talk herself into a total. A component with no evidence behind it is left blank, never scored low — and if those blanks could still lift a company out of the bottom, she parks it at needs more research instead of throwing it away.
What to ask her.
Qualify the researched companies on this campaign.
Why did Löfbergs come out at 61? Which parts pulled it down?
What is missing on the ones you sent back for more research?
Ariadne — finds a person
Her job. For a company that has been qualified, find a named decision-maker — not a shared info@ mailbox — and an address that reaches them, then grade how strong the evidence behind that address is.
How she works. Free evidence first: public pages, free registries, the address pattern other people at that domain use, then asking the receiving mail server whether it would accept the address. Only if all that fails does she ask to spend money on a paid lookup, and only with your approval.
The line she will not cross. She never contacts anyone. No introductions, no test messages, not one. Her instructions are explicit that finding an address and using it are different jobs and only the first is hers.
What she will and will not claim. A mail server accepting an address makes it probable, never verified — because acceptance is not delivery. If you ask her for certainty about an address, she will tell you the strongest thing she can honestly say is that it looks deliverable.
What to ask her.
Find contacts for the qualified companies on this campaign.
Who is the right person at Zoégas — someone in category or brand, not general enquiries?
What is the evidence behind that address?
Talking to any of them
Click an agent anywhere it appears — in your team list, or via Find more companies on a campaign — and you open your ongoing conversation with it. It is one continuous history per agent, not a fresh session each time, so referring back to what you asked last week works.
They are ordinary chat conversations. You can interrupt, argue, ask for the reasoning, tell one to stop, or ask it to explain a decision you disagree with. None of them writes to anybody, so nothing you ask in one of these conversations can reach a prospect by accident.
ready_for_qualification is a bug, not a feature. If one starts speaking in field names, tell it so.The four switched-off seats
The Marketing group also contains four older agents — a Chief Marketing Agent, a Content Writer, an Audience Researcher and a Distribution Manager — which ship switched off. They belong to an earlier, inbound-marketing shape of the product that the outbound campaign machinery replaced. They are left in place rather than deleted, so nothing breaks for anyone already using them, but they play no part in running a campaign. You can ignore them.