How
How Rufus runs a campaign — signal in, approval, send.
For founders, CEOs, and heads of marketing evaluating HITL workflows: here is the rail, a worked example, and what the gate actually checks.
Five-step rail
Connect your stack
Point Rufus at the jobs you already pay for — CRM, prospecting, email, network, booking, and calendar. Bring-your-own-license connectors keep vendors interchangeable.
Hire an outcome
Tell Rufus what you want: raise awareness, get customers, cross-sell, or run a webinar motion. It turns that objective into a coordinated plan across the channels you connected.
Review at the approval gate
Rufus drafts copy, sequences, posts, and connects into a human-in-the-loop queue. You edit voice, claims, and timing before anything leaves your workspace.
Run the supervised motion
Approved work sends through your mailer, publishes to your social accounts, books demos on your calendar, and writes status back to your CRM — one operator seat instead of six tabs.
Measure and compound
Results land on a dashboard you can reconcile with CSV uploads. Edits and outcomes become memory so the next draft starts closer to what you already approved.
One campaign, end to end
1. Signal. A site visitor matches your ICP persona rules — firmographic + behavioral.
2. Draft. Rufus opens an email draft in your voice with a specific ask, not a generic spray.
3. Gate. The item lands in the approval queue: claims, copy QA, rate limits, and send window.
4. You decide. Approve, edit, kill, or pause. Edits become memory for the next draft.
5. Send & compound. Only then does the message leave via your send provider. Outcomes feed the next cycle.
What a gate checks
Claims & copy QA
Unsupported claims and off-voice copy are flagged before you see a green light.
Rate limits
Daily caps and spacing protect domains and LinkedIn safety holds.
Time of day
Send windows keep outreach inside the hours you configured.
Pause switch
Halt everything at once when reputation or capacity requires it.
Approval queue demo
Approve, edit, kill, or pause — same controls operators use in-product.
Memory compounds
Every edit and reject teaches the next draft. Session and workspace learning already reduce rework today.