Build the exception paths before scaling
Start with reliable account and contact matching. Check for existing relationships, opportunities and suppression. Attach the original message and research context so the receiving person can respond without reconstruction.
Automation can propose a category, but ambiguous, sensitive or commercially consequential replies need human review. Record the owner and next action, and make unowned responses visible. Define how the system handles duplicates, forwarded messages and replies arriving after a sequence is paused.
| Decision | Evidence to use | What changes next |
|---|---|---|
| Question, interest or referral | Specific buyer text and account context | Assign a seller or expert and update coverage |
| Timing or objection | Condition, concern or explicit rejection | Record a justified next step or stop |
| Opt-out or delivery issue | Suppression request or technical failure | Stop applicable contact and resolve the error |
Work through the decision
Illustrative mixed reply: 'Not me—ask product, and please remove me from this sequence.' The correct workflow has two actions: suppress the current contact as requested and consider the referral under the appropriate rules. It should not classify the message as positive and continue emailing the original person.
A human reviews the referral context before approaching another role. The CRM preserves the account and reason while limiting access appropriately. Test this mixed case alongside simple interest and simple opt-out; real replies often contain more than one meaning.
An AI category overrides an explicit buyer instruction
If a model says positive while the text requests removal, the explicit instruction must control the contact workflow. Review ambiguous replies and test override rules. Keep buyer text out of general analytics events.
A concrete next step
Test six controlled replies: interest, product question, referral, later timing, opt-out and mixed instruction. Verify record ownership and suppression across connected tools.
Sources and research notes
- FTC: CAN-SPAM business guideRegulator guidance; US scope
- GitLab sales development handbookCompany operating handbook
- Attio: Sharing and permissionsProduct documentation
Primary sources reviewed October 6, 2026. The operating recommendations and worked scenarios are Daavid’s analysis. Illustrative numbers are assumptions, not measured client results. Company marks identify sources and prior experience; they do not imply a customer relationship or endorsement.
Schematic