AI SDRReply classification
Reply classification that says why.
On each thread the AI SDR works, a new reply gets a label, how sure it is and the reason in plain words. The label picks what happens next.
Good timing. We find the people and write a first line for each. Will they work LinkedIn first, or is email your go-to?
AI draft · approved by youLinkedIn first. Sounds useful, can you send a short overview?
Classified asInfo request0.92
Answers the channel question and asks for a short overview.
How it works
From a new reply
to its next step
By default a workflow hands its replies to the AI SDR, so the AI SDR reads every new reply on the workflow's threads.
A reply lands
A lead answers on LinkedIn or by email. Messages sent close together are read as one reply, after a short wait, a minute by default.
It gets a label
The AI SDR reads the reply in its thread and gives it one of 15 labels, such as Interested, Info request, Not now or Out of office, plus one of your own when the reply fits it.
How sure, and why
Each label carries how sure it is, from 0 to 1, and the reason in plain words. The thread keeps both.
The label picks a plan
A reaction, a stage move, an answer through your Playbook, or a hand-off to you. Its scenario sets whether it runs on its own or waits for your OK.
Unsure, it asks you
A reply it would answer but cannot read with confidence, under 0.5, comes to you as a task. So do questions about pricing, a call, legal or security.
What a label does
Each label
runs its own plan
Info request moves the person to Soft Interest, reacts to the message on a first reply and hands the dialog to your Playbook. On a scenario set to Copilot, the plan waits in the thread for your OK, and you can edit its steps.
- Seven labels move the stage, such as Interested to Strong Interest
- Out of office waits until they are back, then follows up
- Do not contact puts the person on the stoplist
- Reacted to their message
- Moved to Soft Interest
- The Playbook, from Discoversent Tue 09:40leadingThe Playbook leads…
EngageDiscoverBridgeOfferBook
Team2 SDRsTheir toolChannelLinkedIn
Wrote: how we help, then the channel question. She named the hires, so it asked how they will reach out.Wrote: the overview, then 15 minutes with Jordan. She asked for an overview, a buying signal, so it went straight to the offer.
Their next reply comes back here. No reply in 3 days: a follow-up with a new angle, 3 at most.
In Unibox
Find replies
by their label
Unibox keeps the LinkedIn and email replies in one list. Its Classification filter lists the labels with a count beside each, so the Interested replies are one pick away.
- The 15 labels in one filter, each with its count
- Pick one, or several, and save the filter
- A reply the AI SDR does not read waits in Needs you
All replies5LinkedIn3Email3
- Maya L.Happy to, Maya. Here is a two-page overview of how teams your size run it with GetSales. Would 15 minutes on Thursday help?Thursday at 2 works. Send the invite to this address.LinkedInEmail·Waiting for replyAI SDR handlingDraft ready
- Lena R.Ask me again after the quarter.LinkedIn·Follow up 12 Jan
- Tomás F.What would this cost for a team of twelve?Email·Handed to you
- Anika B.Thanks. We hired one in August, still ramping up.LinkedIn·AI SDR handling
- Sam K.Which teams in the UK run it today?Email·Your turn to reply
- Interested1
- Meeting request1
- Info request2
- Referral0
- Not now1
- Out of office0
9 more labels
Part of the AI SDR
The first read
of a reply
The label is the AI SDR's first read of a reply. The plan it picks leads to a drafted answer, and each scenario sets whether it sends on its own or waits for your OK.
AI SDR draftWaits for your approval
Happy to, Maya. Here is a two-page overview of how teams your size run it with GetSales. Would 15 minutes on Thursday help?
Approve and sendEditReply as Jordan O.
Questions
Reply classification,
in the AI SDR
Anything else, ask our AI sales agent.
Talk to AI SalesHow do I categorize email responses in sales outreach?
Let the AI SDR label them. By default a workflow hands its replies to the AI SDR, so each new reply on LinkedIn or email gets a label, such as Interested, Info request, Not now or Out of office, with how sure it is and the reason in plain words. In Unibox, the Classification filter shows the replies of one label.
How does reply classification work?
When a lead answers, the AI SDR reads the reply in its thread, with what came before it. Messages sent close together are read as one reply, after a short wait, 60 seconds by default. It gives the reply one label, how sure it is from 0 to 1 and the reason in plain words, and the label picks its plan. A voice message is read once its words are transcribed.
Which labels are there?
Fifteen: Interested, Meeting request, Info request, Referral, Not now, Out of office, Auto-reply, Wrong person, Reschedule, Cancel request, Low signal, Not interested, Objection, Do not contact and Vendor pitch.
Can I add my own labels?
Yes, up to 20 of your own. Tell Creator the label and when it applies, in your words, and the AI SDR reads it in the same read as the 15. A reply gets at most one of yours. You can also have a platform label hand the thread to you instead.
What happens after a reply is labelled?
The label picks its plan, and each scenario sets its level: Autopilot sends on its own, within the caps you set. Copilot drafts and waits for your OK. Conditional reads your conditions first. Seven labels move the pipeline stage, never back down the funnel: Interested to Strong Interest, Info request to Soft Interest, Meeting request to Meeting Requested, Not now to Not Now, Wrong person to Wrong Decision Maker, Not interested to Not Interested and Do not contact to Disqualified. On a first reply, or when the stage moves forward, the plan opens with a reaction: a heart for Interested and Meeting request, a thumbs-up for the other replies a person wrote, and none on Do not contact or Vendor pitch. Out of office waits until the return date plus three days, or 21 days when no date is named, then follows up. Do not contact closes the thread and puts the person on the stoplist at once, at any level. Vendor pitch, a reply that sells you something, closes the thread with no answer. Low signal gets one short answer, then the person goes back to their workflow, whose steps resume.
What if it is not sure about a reply?
When it cannot read a reply it would answer with confidence, under 0.5, it hands the thread to you: a task opens, and the thread notes "Handed to you: the AI SDR could not read the reply with confidence" and waits in Needs you. By default it does the same on a question about pricing, a call request, a legal or security question, a question its knowledge does not answer and a third request to reschedule.
Does it label replies when the AI SDR is off?
No. With a workflow's After a reply set to Stop, I reply myself or to Continue the sequence, on a thread a person switched off, and on an imported old thread, a reply gets no label and no plan and spends no credits: it waits for you in Needs you. A group thread and a message older than the AI SDR's window, 30 days by default, are not read either.
What is AI inbox management?
Software that reads incoming messages and sorts or answers them for you. In GetSales it is the AI SDR working in Unibox, the inbox for your LinkedIn and email replies: it labels each new reply on the threads it works, moves the stage, drafts the answer and hands you the threads it should not answer alone.
Replies read
Each reply it reads has its label.
With how sure the AI SDR is, the reason why and the plan it picked.
- Interested12Strong Interest
- Info request5Soft Interest
- Out of office3After their return
One reply it could not read with confidence came to you as a task.
Every label comes with its reason.