Warming up
Replies from every mailbox and every channel land in one place, tagged by what the person actually said — interested, not now, unsubscribe — and attributed to the campaign that earned them.
The default view is not 'all mail'. It is everyone who replied and has not had an answer. That is the only list that represents work, and it is the one that shrinks as you do the work. An inbox that shows you everything shows you nothing about what to do next.
Each incoming reply is read and labelled — interested, not now, referral, out of office, unsubscribe request. That is what lets a queue of forty replies become a queue of six worth answering this morning, and it is why an out-of-office does not sit in your work list pretending to be a lead.
Every conversation carries the campaign and the sending mailbox it came from. With eight mailboxes across three campaigns, the question 'which sequence is producing the interested replies' is otherwise answerable only by reconstructing it by hand — and it is the question that decides what you write next.
Cold outreach at any scale runs across several mailboxes, and often several channels. The replies, however, are a single stream of work: people who answered a message you sent and are now waiting. Splitting that stream across eight webmail tabs is how interested prospects go cold while sitting in an inbox nobody opened that day.
Sending from multiple mailboxes is not optional — it is how you spread reputation risk and stay within safe per-mailbox volumes. The consequence is that replies arrive scattered across all of them, and the operational answer most teams reach for is a rota of browser tabs.
It fails in a specific and costly way. The mailbox you check least is the one whose warm replies go stale, and there is no signal telling you which mailbox that is. Response speed is one of the few variables in cold outreach with a direct, measurable effect on conversion, and a distributed inbox is a machine for making it worse.
The obvious approach is rules: if the reply contains 'not interested', label it negative. It breaks immediately on real correspondence. 'I'm not the right person, but Sarah handles this' is a referral and one of the most valuable replies you can get. 'Not interested at this price' is a negotiation. 'Interested — but not until Q3' is a follow-up date, not a lead for today.
Warmerly reads the reply and labels it by what it means. The categories are the ones that change what you do next: interested, not now, a referral to someone else, an automated response, or a request to stop. It is not sentiment analysis and it does not produce a score. It sorts a queue.
When somebody writes 'take me off your list' rather than clicking the link, that is an objection to direct marketing and it has to be honoured. Catching those in prose — not only as link clicks — is why intent labelling is a compliance feature as much as a productivity one.
A reply is sent from the mailbox that owns the thread, in that thread, with the correct references and in-reply-to headers. That is not a detail — a reply sent from a different mailbox, or as a fresh message, breaks the thread at the recipient's end and reads exactly like the automation it is.
It also has a deliverability effect. Threaded replies inside an existing conversation are among the best-performing mail you can send, because the recipient has already engaged with the thread. Breaking the threading throws that away.
Replying to a lead also stops the sequence for that person. Nobody who has answered you should receive the automated day-nine follow-up, and it is the single most common way a promising conversation is ruined.
LinkedIn and WhatsApp replies land in the same queue as email, with the same intent labels and the same campaign attribution. A prospect you emailed on Monday and messaged on LinkedIn on Thursday is one conversation with one history, not two disconnected threads in two products.
This is the practical argument for running multi-channel sequences inside one tool rather than stitching two together. The sending is the easy half to replicate; it is the replies that become unmanageable when they are split.
J and K move through the queue, the conversation opens beside the list, and the reply box is where your hands already are. This sounds like a small thing. It is the difference between a tool someone uses for forty replies a day and one they abandon after ten.
The design goal is a queue that reaches zero, every day, in one sitting. Anything that adds a click between reading a reply and answering it works against that, which is why the interface is deliberately sparse.
It is not a CRM. There are no deal stages, no pipeline value, no forecast. Warmerly's job ends when a cold contact becomes a live conversation with a name and a context — after that, the conversation belongs in whatever system your team already runs on.
It is also not your personal mailbox. It syncs the conversations that belong to campaigns; the rest of your mail stays where it is. That boundary is deliberate — a tool that ingests everything is a tool nobody can safely connect a real work mailbox to.
No — only conversations connected to your outreach. Your personal and internal mail stays in your normal client. The boundary is deliberate.
Each reply is read and labelled by intent — interested, not now, referral, automated, or a request to stop. Keyword rules were rejected because they misread the most valuable replies, particularly referrals.
Yes. A lead who has replied is removed from the automated cadence, so nobody who has answered you receives the day-nine follow-up.
Replies always go from the mailbox that owns the thread, in that thread, with correct threading headers — which is both what the recipient expects and better for deliverability.
Yes, in the same queue, with the same intent labels and campaign attribution as email.
One queue of everyone who replied and has not heard back, sorted by what they actually said.