How a sales automation system works, part by part: data, triggers, sequencing, sending, CRM sync and approvals, plus a sample outbound campaign with capacity maths.
A sales automation system is six parts wired together: a data layer holding accounts and contacts, triggers that decide who enters, sequencing rules that pick the next step, sending infrastructure, a CRM sync that records results, and a human approval gate. Each part fails differently, so the system is only as reliable as its weakest link.
Most guides describe a sales automation system as "a CRM plus some rules". That's true for a lead-routing workflow. For outbound, where you contact people who didn't ask to hear from you, the system has more moving parts, and each one can quietly ruin the others. Here's how the parts fit, then a sample campaign wired end to end. If you want the wider picture of rule-based tools versus agents first, read our B2B sales automation overview.
A sales automation system works by moving a contact through six parts in order: data says who exists, triggers decide who enters, sequencing rules choose the next step, sending infrastructure delivers it, a CRM sync records what happened, and a person approves the limits. Activepieces describes a similar three-layer view of data, logic and execution (Activepieces). Outbound needs three more layers than that, because sending and approvals carry real risk.
| Part | Its job | How it usually fails |
|---|---|---|
| Data | Holds accounts, contacts, verified emails, exclusions | Stale or unverified contacts; customers left on the list |
| Triggers | Decide who enters and when | Too broad a filter, or a signal that fires on the wrong people |
| Sequencing | Picks the next step and the stop conditions | Follow-ups sent after a reply or a booked meeting |
| Sending | Authenticated inboxes, caps, unsubscribe handling | One burned domain; complaints above the limit |
| CRM sync | Writes activity and outcomes back, reads exclusions | Two systems overwriting each other's fields |
| Approval gate | A person signs off segment, claims and limits | Nobody reads the queue, so it becomes a rubber stamp |
The rest of this post goes through each part with the rule that matters most.
The data layer needs four things per contact: who they are, a verified address, where they sit in the CRM, and whether they're excluded. Activepieces calls this layer the backbone, and says everything collapses if the inputs are disorganised (Activepieces). We'd add that the exclusion list is the field people forget.
Exclusions come from three places. Existing customers and open deals live in the CRM. Previous unsubscribes and bounces live in the sending system. Competitors, partners and "do not contact" accounts live in somebody's head until you write them down. A system that reads only one of those three will eventually email someone it shouldn't. Our data comes from B2B data provider APIs and is billed by the provider, so the verification step belongs in this layer, before anyone is enrolled.
A trigger starts a workflow, and in outbound there are really only three kinds: a person approving a contact into a campaign, a timer finishing after the last step, and a signal such as a pricing-page visit or a relevant job post. Replies and unsubscribes are triggers too, but they trigger stops, not sends.
The mistake is treating every signal as equal. A visit to your pricing page from a target account deserves a fast, specific follow-up, which we cover in website visitor follow-up. A single blog view doesn't. Decide which signals earn a send, which only raise a score, and which are ignored, and write that down as part of the system. Otherwise the loudest signal wins by default.
Sequencing rules decide the next step by combining a schedule (wait three days) with conditions (no reply, no bounce, not suppressed). The condition check matters more than the schedule. HubSpot's sequences, for instance, unenroll a contact by default when they reply to any email, and when they book a meeting through a link in the sequence, with the option to apply that to the whole company (HubSpot).
That default is the behaviour to copy. Our rule of thumb: check suppression again at the moment of sending, not only when the contact was enrolled. A contact enrolled on Monday can reply on Tuesday, become an open deal on Wednesday, or ask to be removed on Thursday. A system that checked once, at enrolment, sends the next email anyway. The stop conditions we'd always include:
Sending infrastructure has to authenticate every message, spread volume across several inboxes, cap each one, and honour opt-outs fast. These aren't preferences. Google says all senders need SPF or DKIM, bulk senders (more than 5,000 messages a day to Gmail) need SPF, DKIM and DMARC, and spam rates in Postmaster Tools should stay below 0.30% (Google).
| Rule | Source says | What the system should do |
|---|---|---|
| Spam rate | Gmail: below 0.30%. Yahoo: below 0.3% | Auto-pause a campaign when complaints or bounces rise |
| One-click unsubscribe | Google and Yahoo require it for marketing and subscribed mail; RFC 8058 defines the header | Add a List-Unsubscribe-Post header with an HTTPS endpoint, DKIM-signed |
| Honour opt-outs | Yahoo: within 2 days. FTC (CAN-SPAM): within 10 business days | Suppress instantly, across all campaigns, not on a weekly import |
| B2B exemption | FTC: the law makes no exception for business-to-business email | Include a postal address and a working opt-out in every message |
Yahoo's guidance matches Google's on the 0.3% line and adds the two-day opt-out window (Yahoo Sender Hub). RFC 8058 sets out how the one-click header works (IETF). The FTC says each separate email in violation of CAN-SPAM can carry penalties of up to $53,088, and that opt-outs must be honoured within 10 business days (FTC). This is general information, not legal advice. For the deliverability checks in more depth, see deliverability guardrails for AI-sent email, and run your own domain through the email domain checker.
A sales automation system should sync with a CRM in a defined way: it reads accounts, exclusions and deal status in, and writes contacts, activity and meetings out, with one owner per field. StackSync puts it directly: every connection needs its own identity mapping, permitted directions, field ownership and recovery rules (StackSync).
In plain terms: decide that the CRM owns deal stage and account owner, and the sending system owns email status and last-contacted date. If both can edit "lead status", one will overwrite the other and a rep's note disappears. Match contacts on one stable key, usually the email address plus company domain, so a re-import doesn't create duplicates. Our integrations page lists the CRMs, calendars and Slack connections we work with.
A human should approve the campaign file once, before launch: the segment, the claims the emails may make, the sample emails, the sending limits and the stop conditions. After that, people handle exceptions, mainly interested replies and anything touching price, security or legal terms. Approving each email doesn't scale, and reviewers who see hundreds of near-identical drafts stop reading them.
Make the default safe. If nobody answers an approval request, the campaign holds. It never sends. Our post on what an AI SDR should do alone and what needs approval goes through the approval card and the escalation list in detail.
A sample setup is a small written file plus the capacity maths behind it. This one is hypothetical, built from our own assumptions, not a customer result. The target: 600 RevOps leads at 50 to 200 person SaaS companies, three emails each, sent from five warmed inboxes.
segment: RevOps leads, SaaS, 50-200 staff, US exclude: customers, open deals, anyone suppressed sequence: email 1 (day 1), email 2 (day 4), email 3 (day 9) stop on: any reply, booked meeting, bounce, unsubscribe, new deal claims: only facts the Researcher cites from the company's own site inboxes: 5, max 40 emails each per day auto-pause: bounce rate above 2% approval: campaign file and 3 sample emails, then replies by exception
The arithmetic: 600 contacts times three emails is 1,800 emails at most, and fewer in practice, because every reply or booking removes a contact. Five inboxes at 40 a day is a ceiling of 200 a day, so the campaign needs at least nine sending days. That fits inside the 5,000 emails on our Growth plan with room left, and it is more than the 750 free trial emails, which cover about 250 contacts at three emails each. Change the numbers for your own mailboxes in the cold email capacity calculator.
Notice what the person did: wrote the file, approved three samples, and now reads only the replies that matter. Everything between those points runs on the rules.
Build the stop rules and the exclusion list first, then the sequence. A system that sends well but can't stop is worse than no system. Next, get sending authenticated and capped. Only then add triggers and signals, because they multiply volume. If you can't describe your buyer in one sentence, none of the six parts will rescue the campaign.
We run these parts as six agents inside our AI SDR: Researcher, Strategist, Prospector, Writer, Sender and Closer. You approve every campaign before anything sends, in the app, by email or in Slack. You can start the free trial to see the campaign file and approval step on your own segment.
SDR agency vs in house SDR compared on speed to meetings, control, management load, contracts, meeting quality and attrition, with a switch-over checklist.
Eathan Brown · 2 Nov 2026 · 8 min readA Lindy AI alternative for every job: Zapier Agents, n8n, Gumloop, Relevance AI, Apollo, Salesforge, AiSDR and Outbound Labs, with a cost for 500 prospects.
Eathan Brown · 1 Nov 2026 · 9 min readAutonomous SDR defined: what the label means, the autonomy levels from drafts to autopilot, how it differs from an AI SDR or sequencer, and a 4-question test.
Eathan Brown · 31 Oct 2026 · 5 min readA go to market strategy for startups that starts with outbound: when it fits, how to pick the first channel, the capacity maths and a 30/60/90 plan with stop rules.
Eathan Brown · 31 Oct 2026 · 9 min read