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.
How it behaves
Getting here
Opening an agent from AI Agents (its Open button) lands you on this tab. Up to five more sit beside it in the same strip: Leads, Companies, Sources, Agent Instructions and Settings. Dashboard is the default and adds nothing to the address bar; each of the other five writes itself into the URL, as does which settings section is showing, so any of them can be bookmarked or shared. The Dashboard tab itself stays mounted underneath whichever tab you switch to, so its activity feed does not restart just because you looked at Leads and came back. Companies only appears once the agent carries at least one tag: a tagless agent has no company list to show, only a lead pool it shares with everyone else untagged. The header above the numbers always shows the agent’s status pill. The Active/Paused switch, the Pause options menu (Pause everything, Only sourcing, Only outreach) and the Edit settings button appear only for a role that can change agents; for everyone else the header is read-only. See agent statuses and pause options for what each status and each pause choice actually stops.The four numbers at the top
Leads Found and Meetings are current totals, not a rolling window. Leads Found sums the leads carrying the agent’s tags across all six stages. Meetings counts only the leads sitting in the Meeting Scheduled stage right now. One that later moves on to Converted stops being counted here. For a tagless agent both numbers describe the shared untagged pool instead. Contacted and Reply Rate look back over a rolling window instead, each compared against the same-length window right before it. That comparison is the small percentage badge beside the value, with an up or down arrow when it is not flat; the other two cards never show one. Both of those two carry one more restriction: they only resolve for an agent tracked by exactly one tag. Reply Rate is honest about it. It shows a dash in place of the percentage, plus a small info icon that opens a tooltip reading “Reply rate is only available for agents tracked by a single tag.” Contacted carries the identical restriction but no such warning: for a multi-tag or tagless agent it quietly shows 0, which is not the same thing as zero contacted leads.The tab strip’s own signals
Leads, Companies and Sources each carry a small count chip. The Sources chip is just how many lead sources the agent has, and it is absent when the agent has none. The Leads and Companies chips start from the totals the page loaded with, then follow whatever their own table reports once you have opened it, so they climb as sourcing finds more without a reload. Sources and Settings can each also show a small coloured dot. They read the same list of problems the alerts stack renders, but each takes only its own slice: Settings takes the two blocking ones (no sending channel connected, and a goal with no call-to-action), so its dot is always red. Sources takes the two sourcing ones (out of credits, and the company limit reached), so its dot is red when credits ran out and orange when only the limit was hit. Every other problem raises a banner without lighting any tab. The Settings dot (1) sits immediately after the tab’s label. A role that can change agents also sees an Add leads button at the right edge of this strip, once the agent has a tag to add leads into.The alerts stack
Inside the Dashboard tab, above the two columns, sits a stack of coloured cards listing whatever is actually wrong with this agent: one card per problem, each with a button that either does the fix in place or takes you to where it is done. The order is fixed by problem type, not by severity, so a warning can sit above an error. Nothing renders at all when the agent has none. See agent alerts and what to do about them for what each banner means.The main stage: recent activity
Below the alerts, the wide column is a single merged, newest-first timeline of what the agent has actually done (an email sent, a reply received, a new lead found, a company analyzed), capped to a fixed number of the most recent items and refreshing itself quietly while you stay on this tab. A tagless agent gets only the email rows, because there is no tag to find its leads and companies by. A placeholder message replaces the list until the agent has done something. An email row (1) opens its full thread in a dialog without leaving the page; a lead or company row takes you to its own record instead. Open full activity monitor (2) leaves the agent entirely for the workspace-wide Activity Monitor.The right rail
Three cards, in this order. Sending limits comes first. It shows one row per connected channel, and the headline ”≈” figure is this agent’s share of that channel’s allowance for today. The line under it says how much of that share is already spent and how much is left, or, when nothing is going out, why: still warming up, today’s limit reached, or nothing available to send. Everything forward-looking carries an approximation mark because the underlying limits are recalculated daily and split across every active campaign using the same account; see why the agent’s daily sending numbers move for the mechanism. With no channel connected at all the card says so and offers Add a channel instead. It disappears from the rail entirely in one case: channels are connected but the agent’s data carries no capacity figures, in which case the card removes itself rather than show numbers it does not have. Lead funnel sits below it. It draws the agent’s six stages as horizontal bars in pipeline order, from Untouched to Closed, each bar’s width measured against the largest stage and the count on the right. Clicking a bar jumps straight to the Leads tab already filtered to that stage. See lead stages and the funnel view for what actually moves a lead from one stage to the next. Setup checklist is last. It tracks the handful of things a working agent needs; it collapses to just its header the first time every row is done, and remembers whether you left it open or closed separately for each agent. See the agent setup checklist explained for what completes each row.Leads and Companies show the same slice, filtered
Opening Leads or Companies shows a table scoped to the agent’s tags, with a search box and, once the agent has at least one lead source, a per-source dropdown. That source filter is shared between the two tabs, so picking one source on Leads keeps it selected if you switch to Companies. Only the Leads table adds stage chips across the top; the Companies table opens with a count of the companies matched instead. A lead still being enriched keeps its usual row but dims its photo and spins a loader over it; a company still finding its people shows Finding leads… in its Leads column, where the row of faces it ends up with would go. Both tables poll themselves while any visible row is still mid-analysis (Companies keeps polling while a row is still finding people too), and stop once nothing is.While a lead source is still starting up
A lead source shows Pending before its first run. The dashboard re-checks the agent by itself for a few minutes after you land here, so a source that starts server-side flips to running without a manual refresh. A source that is waiting for you to start it stays Pending, and the polling gives up rather than waiting forever.Limits
Related
Why the agent's daily sending numbers move
The full mechanism behind the Sending limits card’s ”≈” figures.
Agent alerts and what to do about them
Every banner the alerts stack can raise, and the one action that clears each.
The agent setup checklist explained
What exactly completes each row on the Setup checklist card.
Lead stages and the funnel view
The six fixed stages the funnel card draws, and what moves a lead between them.