ForgeSendDocs

Guided actions and confirmations

Beyond answering questions, the AI Account Analyst can take six kinds of action directly — always shown as a preview first, and never run without an explicit Confirm click.

The six guided actions

ActionWhat it does
Pause a campaignStops an ACTIVE campaign from sending further steps
Resume a campaignRestarts a PAUSED campaign — re-runs the full preflight check as part of resuming, and can be blocked if something changed while it was paused
Reassign sender inboxesSets a campaign's sender pool to a specific list of inboxes — only while the campaign is DRAFT or PAUSED
Tag a replyApplies an intent tag — Interested, Not Interested, Out of Office, or Unsubscribe — to a reply in Reply Box
Mark a reply as readMarks a single reply read
Pause or resume a campaign in another client workspaceSame as pause/resume above, but for a campaign in a different client workspace under the same agency — only available to an Agency Group owner
Clone a campaign into another client workspaceDuplicates a campaign from the current workspace into a different client workspace under the same agency — the clone starts as an empty DRAFT with no leads or sender inboxes carried over

Ask for any of these in plain language — "pause my Q3 campaign," "mark that last reply as read" — and the Analyst resolves the campaign or reply you mean itself before proposing anything.

Pausing/resuming or cloning a campaign in another client workspace is only available if you own an Agency Group — and only for a client workspace you're also a member of, not just one your agency owns. You refer to the target client by its exact workspace name rather than an ID; the Analyst looks up your agency's real client workspace names first if it doesn't already know them. Every preview for one of these states both the current workspace's name and the target client workspace's name plainly, so it's always clear which client's data is about to change.

How a confirmation works

Every guided action is a two-step call under the hood: a preview that makes no changes, and an execute that actually runs it. The Analyst always shows the preview and waits for you to click a real Confirm button — it never infers confirmation from wording like "yes" or "go ahead," no matter how clear it sounds.

Guided action preview — read only

Preview — pause campaign

"Q3 Outbound — SaaS VPs" is currently ACTIVE. Pausing it stops any further steps from sending; leads already enrolled stay enrolled and can be resumed later.

ConfirmWalk me through it

Each preview mints a single confirmation that's good for 15 minutes and can only ever be used once — clicking Confirm consumes it immediately, so the same action can never accidentally replay later, and an old preview left sitting on screen simply goes stale. If the conversation moves on, even a new message that seems to reference the same pending action triggers a fresh preview rather than reusing the old one.

Confirmed actions also share a workspace-wide rate limit — 20 AI-initiated writes per rolling hour, counted across guided actions and Launch Guardian together. Hitting it returns a plain message telling you roughly how long until the next slot frees up, rather than silently failing.

"Walk me through it"

For any action with real consequences, choosing Walk me through it instead of Confirm guides you through doing it yourself in the normal interface, step by step, rather than the Analyst doing it automatically. Lower-risk actions — pause, resume, mark as read — show Confirm as the visually primary option, since a wrong or hasty click there is easily reversible. Reassigning senders and tagging a reply give Walk me through it the same visual weight as Confirm instead, so the one-click automated path is never the obviously-easiest choice for something with a bigger blast radius.

Tagging a reply: the cascading effects

Tagging a reply Interested or Unsubscribe doesn't just label it — both pause the lead's remaining sequence and skip every pending scheduled send for that contact on that campaign. Unsubscribe goes further and adds the contact to the workspace's suppression list, blocking them from every campaign in the workspace, not just this one. The preview always states these effects plainly before you're asked to confirm — they're never left implicit. Not Interested and Out of Office tags are informational only and don't trigger either effect; see Auto-pause and bounce handling for the full mechanics either tag can trigger.

Auto-tagging out-of-office replies

One narrower permission works differently from everything else on this page: auto-tag out-of-office replies lets the Analyst tag an obviously-OOO reply on its own, with no confirmation step at all — off by default, and only available as an opt-in toggle. It only ever fires when the AI's own classification of the reply is genuinely high-confidence OOO, and only on a reply that doesn't already have a tag, so it can never override a tag you or the Analyst already applied. This is separate from asking the Analyst to tag a reply OOO yourself through chat — that still goes through the normal preview-and-confirm flow above using the Tag replies permission, regardless of whether auto-tagging is on.

Turning actions off for yourself

Settings → Account has a grid of toggles, one per action type, letting you turn any of them off for yourself specifically:

Settings → Account — AI actions — read only
Pause campaigns
Resume campaigns
Reassign sender inboxes
Tag replies
Mark replies as read
Pause suppressed leads
Re-verify campaign leads
Auto-tag out-of-office replies

The first five map directly to the guided actions above; Pause suppressed leads and Re-verify campaign leads gate two of Launch Guardian's own auto-fixes independently of its third (sender reassignment, which reuses the Reassign sender inboxes toggle). Every toggle here defaults on except Auto-tag out-of-office replies, which defaults off — it's the only action that ever runs with no confirmation, so it has to be opted into rather than opted out of. Turning one off doesn't disable the Analyst answering questions or previewing that action — it can still show you what it would do, but attempting to confirm it tells you plainly that you've turned it off for yourself and how to turn it back on.

These preferences are personal, not per-workspace — they follow you across every workspace you're a member of, and don't affect what any other team member's own toggles are set to.

Where to go next

Next: Launch Guardian