Skip to content
Current/ Help Center

AI sequences and campaigns: drafted by AI, sent by you

6 min read · Updated Jul 17, 2026

Two different tools sit behind one idea. A sequence is a paced, personal, one-to-one walk through several touches — email, wait, follow-up task, a condition that checks for a reply — sent from a seller's own mailbox. A campaign is a one-to-many send to an audience, at once. Current's AI can draft both. Neither one sends a single message you haven't approved.

What the AI actually does

You describe the goal in plain words — "revive quiet managed partners into a business review" — plus an optional audience hint and a tone (professional, friendly, or direct). The AI drafts a complete multi-step sequence: three to six emails with sensible spacing, follow-up tasks, and a no-reply condition, shaped exactly like the steps you'd have built by hand, so the draft drops straight into the builder for you to edit. For a campaign, the same thing happens for a single email: subject, preheader, and body blocks.

The drafting tools write nothing and send nothing. They hand you a draft and stop. And they're validated on the way out: the AI cannot invent a link (any web address it produces that wasn't in what you gave it is stripped before you see the draft), and everything it writes is length-capped. On the sequence side the draft is also merge-tag-scrubbed — any tag Current can't actually resolve for a sequence send is removed. On the campaign side an unknown {{token}} can still appear in the draft you review, but the send preflight refuses to queue the campaign until you fix it, so it can never reach a recipient. What you paste in is treated as material to write from, never as instructions to obey.

  1. 1
    Open Sales → Sequences and click Build with AI
    Describe the goal, optionally who it's for, and pick a tone.
  2. 2
    Read the draft properly
    It arrives as real, editable steps in the builder — not a blob of text. Rewrite the subject lines that don't sound like you, delete the touch that's one too many, change a wait from three days to five.
  3. 3
    Save it, then turn it on
    Nothing is written until you click Save as paused sequence — and it saves paused, so no one is enrolled and nothing is queued until you activate it. See "How to build and launch a sequence, step by step" for the full walkthrough.

The guardrails

  • No AI-drafted email sends until you've read it and saved it — the drafting tools write nothing and send nothing.
  • Sequence emails go out from the seller's own Microsoft 365 mailbox — a real person's address, not a shared blast domain — and land in their Sent Items.
  • Business days only, on by default per sequence: a step falling on a weekend or one of your holidays re-arms to the next business morning. Because that's applied when the step is due, editing your holiday calendar affects sends that are already queued.
  • A hard cap of 250 sequence emails per mailbox per day. Number 251 isn't an error — it reschedules to the next business morning and is logged as such.
  • A reply takes the contact out of the sequence before the next touch can go — the reply sweep runs ahead of the send sweep on every cycle (a few minutes end to end), so a replying contact doesn't get one more automated email.
  • Bounced and unsubscribed addresses are suppressed. Campaign sends re-check suppression at the moment of sending. Sequences check at enrolment and again on every sweep — the sweep runs before any due email goes out, so a contact who unsubscribes mid-sequence is pulled out before the next touch.
  • Email verification, when NeverBounce is connected, keeps invalid addresses out of the send entirely. See "NeverBounce: verify email deliverability before you send."

The full mechanics — auto-enrolment triggers, dedupe rules, every exit condition, and what happens if you delete a sequence with people still in it — are in "The sequence engine: enrollment, caps, and exits."

Heads up
Campaign sending has to be armed deliberately
Until your sending domain is configured and live sending is explicitly switched on, campaign sends run in preflight: anything not on the safety allowlist is refused rather than delivered. This is intentional. It means a misconfigured workspace cannot accidentally email four thousand partners on its first afternoon. See "How Current sends email — and why two mail paths, not one."

Targeting: audiences are live queries, not static lists

An audience is a live query over your PSA data, not a spreadsheet you exported last month. Filters stack, and every one of them is answered from data that syncs on its own — so an audience built today is still correct in six weeks. You can filter by:

  • Relationship — all, customers, or prospects.
  • Account owner — sales leadership can pick any account executive or "Mine"; an account executive's audience is pinned to their own accounts.
  • Location (state and city), industry, and classification.
  • Contract type, and companies holding an active service line by name — the basis for a price-increase send.
  • Offerings held: has any, has all, or has only — plus "missing these services," which is how you target a whitespace gap.
  • Tags, and saved lists — including the lists that arrived with your HubSpot import. You can add a static list to the audience, or load a dynamic list's stored filters.
  • Minimum MRR (a financial filter, so sales leadership only).
  • Contacts — primary contact only, active contacts only.
  • Deliverability — optionally drop catch-all or unknown addresses from this send.

A useful worked example: notifying every partner on a Managed Services agreement of a price change takes about a minute to target — pick the service line, and Current resolves the companies, their contacts, and each one's current price. Save the filter set as a named segment and it's reusable forever. See "Campaigns and audiences: build, approve, and send safely" and "How to send your first email campaign, start to finish."

Watching what it does

Each sequence has its own report: a full-screen funnel of emails sent, replies, meetings booked, exit reasons, and exactly where every enrolled contact currently sits — with replies as the headline metric, because replies are the only metric that pays. Sequence email leaves from the seller's own mailbox, so it isn't open-tracked; open rates are a campaign number, not a sequence one. A reply also rings the enrolling seller's bell the moment it's detected, which is the one notification nobody has ever asked us to turn off.

Tip
Read the exits, not just the sends
The interesting numbers are replies and exit reasons. A sequence where most people exit as "completed" without a single reply isn't a sequence with a deliverability problem — it's a sequence with a message problem. Rewrite the first email before you touch anything else.
Note
Every send is a real activity
A sequence email is logged on the contact and company timeline like any other touch, and a sequence task lands in the seller's to-dos with the company and deal attached. The AI's outreach doesn't live in a parallel universe — it shows up in the same history as the call you made yesterday.
Was this helpful?