Email & sequences
Email and sequences that can't send without consent
1:1 email, campaigns, follow-up sequences and draft-and-approve growth agents — every send resolved against consent at one chokepoint, on your own provider keys if you like. Currently in beta.
What it solves
Outreach tools make it easy to blast a list and hard to stay compliant while you do it. The moment sending and consent live in different places, a mistake is one careless campaign away. Finocket routes every message — 1:1, campaign or sequence — through a single chokepoint that checks consent, suppression and jurisdiction rules before it can leave, so compliance isn't a discipline you have to remember.
How we're different
There's exactly one way a message leaves, and it's consent-checked — no side door. Blocked sends are logged with a reason rather than silently dropped, sequences stop the instant someone unsubscribes, and a lint gate flags marketing dressed up as transactional. You can also bring your own email, WhatsApp or SMS provider keys so sends go out on your billing — and the same chokepoint still decides what's allowed to leave. Even the autonomous part is bounded the same way: a growth agent may draft a whole campaign, but it structurally cannot send one. It's in beta today, via early access rather than general availability.
Inside Email & sequences
One send chokepoint
Campaigns & follow-up sequences
One-click unsubscribe
Email & sequences
One-click unsubscribe
Bring your own sending accounts
Email & sequences
Bring your own sending accounts
Consent-checked segments
Email & sequences
Consent-checked segments
Email & sequences
One-click unsubscribe
Email & sequences
Bring your own sending accounts
Email & sequences
Consent-checked segments
More in Email & sequences
A ladder that cannot double-send
Sequences own no transport of their own — they ride the same consent-checked send path as everything else, and neither advancing a ladder nor recording a send is possible without a chokepoint verdict in hand. Exactly-once is the database's guarantee rather than the application's: a unique constraint per enrolment and step means a crashed cron tick that retries cannot send the same rung twice, and a duplicate enrolment of the same contact is refused the same way.
AI-drafted ladders arrive disarmed
A follow-up ladder the assistant drafted is installed switched OFF, and enrolling anyone into a disarmed ladder is refused on the server, not just greyed out in the screen. A person arms it. Every scheduled step is also re-linted for content at send time, so a template edited after approval is still checked before it leaves.
A growth agent that drafts and stops
Let an agent pick an audience from a saved segment, draft the copy and propose the timing — then stop. It writes one proposal and waits. It cannot send, because it has nothing to send with: no transport, no queue and no enrolment writer exist anywhere in it. Your approval binds to the exact copy you read — the plan is hashed when written and the database refuses any later edit to it — and a second tap on Approve is a no-op rather than a second campaign. Undo stops the REMAINING steps; it never claims to recall a message already sent, and no consent verdict is cached, so someone who opts out after approval is still stopped at their next rung.
An agency seat that can't read your list
Give an outside marketing partner a time-limited seat scoped to one campaign and a capped audience. They may draft only — never send, never approve, and never read your contacts, message log, enrolments or templates, because row-level security means there is no request that returns them. The person who proposed a campaign can never be the one who approves it, and a refused attempt is written onto your own AI action log so you see that it happened. The grant log is append-only: revoking is a new row, never a quiet edit.
Next-best offers, with no AI in the decision
Two deterministic detectors over your own invoice lines: a replenishment nudge on the MEDIAN gap between a customer's purchases (never the mean, so one festival bulk order can't hide a six-week rhythm), and a cross-sell suggestion from what your customers actually buy together, measured per customer basket rather than per invoice. Every suggestion carries evidence you can check against your own records. It produces offers and nothing else — there is no message-emission path in it — and the reasoning is counts only, so no other customer's name can appear. Too little history yields zero offers plus a note on what's missing, never a padded list.
Works with the rest of Finocket
Part of Outreach & compliance. See how every module connects.
Questions
Can a message go out without a consent check?
No — every message, whether 1:1, a campaign, a sequence step or an AI-drafted reply, passes through one chokepoint that resolves it against consent, suppression and jurisdiction rules per recipient at the moment of sending. A blocked send is logged with its reason AND keeps the exact copy that would have gone out, so it's auditable rather than silently dropped.
Can I send on my own WhatsApp, SMS or email account?
Yes — connect your own provider keys and sends go out on your accounts and billing, stored encrypted and never shown again. The same chokepoint still decides what's allowed to leave: consent and the suppression list are re-checked per recipient at the moment of sending, whoever's transport carries it.
What happens if the sequence runner crashes mid-run?
Nothing is sent twice. Exactly-once is enforced by a unique constraint in the database per enrolment and step, with the runner claiming a step before it sends — so a retried tick after a crash is a no-op rather than a second email. A duplicate enrolment of the same contact is refused the same way.
Can I target a saved segment?
Yes — save an audience from tags, consent state or activity and send a campaign to just that segment. A segment only SELECTS who's in scope; the send chokepoint still DECIDES each message, so a non-consenting contact inside a segment is blocked, never emailed.
How autonomous can a campaign get?
Autonomous in the thinking, never in the sending. A growth agent picks an audience, drafts the copy and proposes timing, then writes one proposal and stops — it owns no transport, no queue and no enrolment writer, so it has nothing to send with. Your approval binds to the exact copy you reviewed: the plan is hashed at write time and the database refuses any later edit, and a double tap on Approve is a no-op. Undo stops the remaining steps and never claims to recall a message already sent.
Can I let an agency run campaigns without giving them my customer list?
Yes. An outside partner gets a time-limited seat scoped to one campaign and a capped audience, and may draft only — they can never send, never approve, and never read your contacts, message log, enrolments or templates, because row-level security means no request returns them. The proposer can never be the approver, a refused attempt is written onto your own AI action log, and revoking the seat is a new row in an append-only grant log.
Is this generally available?
Yes — Email and sequences are live and generally available.
Learn how it works
Growth agents — campaigns that wait for you
An agent picks the audience, writes the messages and works out the timing, then stops. Nothing reaches a customer until you have read the whole plan and confirmed it, and every message re-checks that person's consent at the moment it goes.
8 min read
Email, campaigns and consent
Launch email campaigns with audience counts up front while one send chokepoint applies consent, suppression, your own sending hours and your country's rules to every single message — blocked sends are logged, never silently dropped.
Email sequences (automatic follow-ups)
Build a day-offset email drip — enrol a contact once and each step sends automatically on schedule, with every step consent-checked and the whole sequence stopping instantly when someone unsubscribes.
Ready to ditch the spreadsheet?
Free for solo users. No credit card. Your books stay yours.