Needle - The proactive GTM agent in Slack and Teams
by•
Most sales AI waits for you to ask. Needle is proactive. It works like a GTM engineer on your team: it watches your pipeline and acts before you do. Spots stalled deals and drafts the follow-up, preps you before calls, keeps your CRM tidy, surfaces real buying signals. It lives in Slack and Teams, wired into HubSpot, Gmail and Gong. Unlike horizontal agents, it is built for revenue teams, acts through your permissions, and your context and memory stay portable. No lock-in. Not another dashboard.
Replies
Very interesting one! Wondering how does Needle tell a deal that stalled from one that went quiet on purpose? Is it pulling that from CRM notes? Call transcripts?
Needle
@artstavenka1 good question, and it's genuinely the hard part. Needle pulls from everything connected, CRM signals like stage and activity, but also email threads, Slack, and yes, call transcripts via Gong and Fireflies. That's what makes the distinction possible: a deal where the champion said "circle back after our board meeting in July" on a call reads completely differently from one that just went dark, even if both look identical in the CRM timeline. The more sources connected, the better it gets at telling stalled from quiet-on-purpose. Still not perfect, edge cases exist, but the whole point is judging from full context, not just a "no activity in X days" timer.
Needle
super hyped to build this!
Needle
@onur_o, proud of you and the team man. 🙌
The Slack/Teams angle makes a lot of sense to me. From my perspective, a lot of GTM tools end up living outside the flow of work, so people either forget to use them or only check them when something has already gone cold.
How would you avoid this becoming noisy though? If Needle is proactive, how do you decide what is worth surfacing versus what just becomes another notification people start ignoring?
Needle
@tobiasfleischer honestly, this is the thing that worries me most too. Right now we handle it two ways: every ping has to come with something the rep can actually do, not just an observation, if there's no decision attached we don't surface it. Notification fatigue is a real risk with anything proactive, and we're still learning where the line is as more people actually use it day to day.
@jan_heimes Yeah, that make sense and good to hear that you're treating notification fatigue as a real risk. Appreciate the answer.
Needle
Hyped!
Built our GTM from scratch using Needle.
Now we have superpowers 🦸
LFGO 🚀
Needle
@mauricevv LFGO indeed 🚀 happy to build this with you. Superpowers unlocked 🦸
Proactive feels right, it pinged me about a stalled deal before I even opened HubSpot that morning and drafted a solid follow-up. The Slack-first setup keeps it out of my way.
Needle
@erolsalmaney2h This is exactly the moment we built it for, catching the thing before you even go looking for it. Thanks for your comment!
Needle
@prachi_nagwan thank you! we believe it's super helpful too.
Congrats on the launch! The Slack and Teams-first angle makes a lot of sense here, especially for the “act before someone opens the CRM” moments.
Curious how you handle trust when Needle suggests an action. Do reps get a clear reason for why something was surfaced, like the signal, source, and suggested next step, or is the goal to keep it mostly invisible unless they ask?
Needle
@akashnawani not invisible, that would make it impossible to trust. Every suggestion comes with why it's showing up, what signal triggered it, where that signal came from, and what the suggested next step is. No black box nudges. The goal is you can glance at it and immediately know whether to act, snooze, or ignore, not have to go dig for context first. If it can't explain itself, we don't ship the nudge.
A proactive GTM agent only works if the handoff is legible. The important part is not that it finds every possible signal; it is that sales/support can see why this account matters, what changed, and what action is safe to take next.
Needle
@krekeltronics couldn't agree more, and you've put it better than most of our internal docs. Finding signals is honestly the easier half, models are good at that now. The hard part is exactly what you said: making the handoff legible enough that a rep trusts it in three seconds.
Interesting approach 🙂. The 'mirrors your permissions' model avoids a lot of the config headache. Curious how it handles document-heavy steps (proposals, contracts, PDFs) in the pipeline. That's often the messiest part of automating a sales workflow.
Needle
@benjouss
Great question, and you're right that it's where most sales automation quietly breaks. Today Needle reads the docs already living in the deal (PDF proposals, email attachments, HubSpot files via Gmail/Drive/HubSpot) and treats them as context: it pulls the key terms, flags what's missing or inconsistent, and drafts the follow-up or proposal off the actual document instead of a generic template.
What it won't do yet is generate a fully redlined contract end to end. Curious which doc step is messiest for your team, proposals or contracts?
@jan_heimes I work on the document infrastructure side (PDF/OCR/e-signature APIs), so I was asking more from that angle.
What's your read on it internally, is it more that contracts pile on complexity (redlining, e-signature compliance), or that proposals are just as messy but higher volume?
the part that gives me pause is "tidies the CRM on its own" - CRM data being wrong silently is worse than it being stale, because nobody double checks a field that looks filled in. is there an audit trail showing what it changed and why, or do you just have to trust the drafts before they go out
Needle
@omri_ben_shoham1
The goal is that "tidy" never means "confidently wrong." Can start with propose only and write only if agreed / review. Once you trust after some time to you can let it run automatically end to end. Also has memory so it learns over time.
propose-only as the default makes sense to me, that's basically the tradeoff every agent tool ends up making eventually. curious how long people usually stay in review mode before flipping to auto - is it a fixed number of clean runs or fully up to the user
memory learning over time is the part that actually sells me on it, most propose-only tools just stay propose-only forever because nobody bothers to loosen the leash. is the trust threshold something you tune per workflow or does it apply globally once it kicks in?