Skip to main content
An agent does not find leads in one move. It searches for companies, scores them against your profile, picks the decision makers inside them, then verifies their addresses. So the first leads arrive gradually. The card that narrates the wait is not measuring any of that work.
Where
Outreach→AI Agents
Your role needs
Read access to AI Agents read-campaigns. Admin, Member and Viewer have it by default.
To create or change
create-campaigns to add one, update-campaigns to change one, on top of the permission above.
If you cannot find this in your sidebar, your workspace may have a custom menu configuration. Contact support and we will check it for you.
The narration (1) is a script. The counters (2) are queries against your own data. Only the second is measuring anything, and even that measures your tag rather than your sources.

​
How it behaves

​
The card only exists for a short window after a launch

When you create an agent yourself, the last thing the wizard does is open that agent’s Dashboard tab and record the moment you launched it. The first-leads card is drawn from that record, so it appears only on the agent you just launched, only in the browser tab you launched it from, and only for a while afterwards. Reloading the page does not restart the window. It carries on running, and once it closes the card does not come back. The agent built for you during onboarding never shows this card, because nobody launched it from the wizard. That is not a fault: everything the card reports is also on the agent’s Sources tab and in the Prospects list, which is where you watch progress once the card is gone.

​
The narration is a script, not a progress bar

While the card is up, it narrates the sourcing pipeline one line at a time, in this order:
  • “Searching matching companies…” When the source has an estimate or a company cap to go on, the same line instead names roughly how many companies the search covers.
  • “Scoring matches against your profile…”
  • “Finding decision makers…”
  • “Verifying contact emails…”
  • “Preparing first outreach…”
The lines advance on a fixed timer from the moment the card appears, and the last one stays up indefinitely. Nothing on that ticker is fed by the agent. It does not know whether a single company has been analyzed, and it never announces that the work is finished. The same ticker runs wherever else the product makes you wait: while an email account’s connection is being tested, while a payment confirms, while an import chews through its rows. It is there to make a wait legible, not to report on one.

​
The two counters are the only real measurement

Under the narration sit two pills (companies found and leads found) and a row of avatars of the newest leads. Those are counted, not animated. They are polled on a fixed interval rather than pushed, and the interval is long enough to notice. A pill that has not moved since you last looked may only be waiting for its next refresh. What they count is the agent’s tag, and nothing else. Both queries ask how many companies and how many leads carry that tag, the same question the Companies tab asks. That is close to “what this agent’s sources have produced”, but it is not the same thing:
  • Leads you add yourself through Add leads land on the agent’s tag, so they are counted here alongside the sourced ones.
  • Anything a source produced without the tag is not counted, and neither pill knows which source anything came from.
A count that fails is silent. The narration keeps playing, the number does not move, and the next attempt usually puts it right.

​
An agent with no tag counts nothing

An agent finds its leads through its tag, and both counters are scoped by it. An agent with no tag has nothing to count with: both pills read zero and no avatars appear, whatever else is going on. The agent’s Companies tab is absent for the same reason. Check that first, before you conclude nothing is happening. Zeros that never move look identical to a stalled search, and as the next section explains, this particular pair of zeros never graduates to a warning.

​
”Sourcing is queued” needs a count that ran and came back zero

When a company count has actually come back, and come back zero, and enough time has passed since the launch, the narration is replaced by a Sourcing is queued — this can take a while heading, warning that this can take a while, with “Your agent will start filling this page as soon as its search reaches the front of the queue.” underneath. All of that has to hold, which makes the message narrower than it looks. One company found is enough to keep the narration running, however slow the rest of it feels, so it is not a general slow-progress warning. More importantly, a zero on the pill is not the same as a count of zero. The pill has no way to say “not counted”, so it draws a zero whether the count came back zero, no count has run yet, or every attempt has failed. Only the first of those can produce the queued message. An agent with no tag never asks for a count at all, so it sits on “0 companies found” and never reaches the queued message, however long you wait.
A card stuck on “0 companies found” with the narration still playing tells you nothing on its own. Check that the agent has a tag, then read the Sources tab, which reports what the sources are actually doing.

​
The green card is the first batch, not the finish

Once enough leads carry the agent’s tag, the narration is replaced by a green card. It names how many leads the agent found and keeps the same avatars. Its See leads button switches to the agent’s Leads tab without leaving the agent. The card says so itself: sourcing keeps running in the background, and it invites you to review the leads while outreach ramps up. Nothing about this card means the source is done.

​
After the card, the Sources tab is where you watch

Every source feeding the agent is one row on its Sources tab, with three numbers: Companies (analyzed so far, against the total cap), Daily (companies per day) and Leads found so far. The status chip on the row is what tells you whether it is working. Pending means exactly that: the source has not started, and its tooltip says to press the start button. What gets it out of Pending depends on which agent it belongs to, and the two paths are not equally patient. The source under the agent built for you during onboarding starts itself. A background job waits for the searches behind it to be built, flips the source to Active as soon as they exist, and keeps re-checking until it runs out of tries. If the searches never arrive before then, the source is left Pending rather than failing. A source you create yourself, in the creation wizard, or by adding one to an agent later, gets one attempt, not a job that waits. The moment it is created the app asks for it to be started, and that request is refused if the searches behind it do not exist yet. The refusal is silent and nothing tries again. The source stays Pending until you press the play button on its row. So a source that is still Pending long after you built the agent may not be starting at all. It may have missed its only attempt. While any source on the agent is Pending, the agent’s page re-checks for a few minutes after you open it, so a source that does start flips to Active in front of you without a refresh. A source you paused yourself is never restarted for you. Two things about that row are easy to misread:
  • A source that has used up today’s allowance still shows the green Active chip, and its tooltip says the daily limit was reached and it will resume automatically tomorrow. It is not a fault and there is nothing to restart.
  • The play button is only enabled for a source that is Pending or that somebody paused by hand. A source stopped by running out of credits, or by reaching its company cap, has to have that cause fixed first. The tooltip says so, and a source at its cap gets an Increase limit link next to the button.
Pausing, resuming and editing either cap all need permission to change lead sources. A read-only role sees the same numbers with none of the controls.

​
Where the leads themselves land

Leads keep arriving whether or not any card is up. From inside the agent, See leads opens its own Leads tab; for the workspace as a whole they collect in the Prospects list under .

​
Limits

The agent setup checklist explained

The card next to this one, and the two rows only working time can tick.

Lead sources: add, edit limits, pause

Every status a source row can show, and what changes each one.

Working the Prospects list

Where the leads land, and how to read them once they do.

How many credits does a lead cost?

Why finding people is metered, and what each one spends.