A team of AI agents that get your buyer moving.
Start with one. It writes your call notes, drafts your follow-ups, or opens accounts that have never replied. Every one of them is judged on the same thing: did the buyer actually move?
- One job each
- One connector to start
- Runs in Slack — no new app
No demo call. No philosophy. Pick the job you hate most.
Fireflies.aiReads the tools you already use. Nothing new to learn.
Gmail
Outlook
Google Calendar
Google Meet
Zoom
Microsoft Teams
Slack
WhatsApp
HubSpot
Salesforce
LinkedIn
Claude
ChatGPTOne team, four moments that matter.
From the first briefing to the forecast, each agent owns one moment and writes back to the same Deal Memory. Start with one; add the rest when you need them.
Before the call
Walk in ready
Meeting context, changes, risks, and the question worth asking.
After the call
Turn the meeting into motion
A clean record and a next step drafted while the conversation is fresh.
When momentum stalls
Re-open the path
Find the quiet thread, the missing person, and the shared place that moves consensus forward.
Across the pipeline
Keep every deal honest
Buyer evidence shapes methodology, forecast, and coaching—not seller activity alone.
Where they work
They post where you already work: approve, edit or send from the same Slack message, or run the same agents as MCP tools inside Claude and ChatGPT. No new app, and nothing is gated by channel.
Every signal and every gap we track is the buyer.
Not your activity — theirs. What they said, who they named, what they accepted, what they never replied to. Each one is scored, turned into a gap, into an action you approve, and then measured to see whether the buyer actually moved.
- Signalsomething happened
- Gapsomething is missing
- Actionyou approve it
- Measureddid the buyer move
- Fireflies.aiTue 11:04Signal
The buyer names someone who is not in the room
I'll need to pull my manager in on the budget.
A mention, not an introduction. The model logs it as medium evidence of a missing stakeholder — the kind of line that dies in a call recap.
NotetakerSIG-BUY-007EV-015 · medium - Google CalendarTue 11:31Signal
The next meeting is accepted
Thursday 14:00, accepted against a stated purpose. Strong evidence — and the deadline every step after it is due against.
SIG-BUY-002EV-002 · strong - LinkedInTue 11:31Gap found
The manager is missing from the invite
Attendees read against the mention: nobody with authority is on it. Nudge maps him — Karim Bensalah, VP Finance — and reads the role off his profile rather than guessing.
StakeholderGAP-EV-003CFG-009MR-009 - SlackTue 11:32Action · proposed
“Manager is missing from Thursday's meeting”
His profile attached, the ask already drafted, due before Thursday. You accept it in the thread — no new app, no dashboard to remember to open.
linkedin.com/in/karim-bensalahACT-EX-001PROPOSED → ACCEPTED - WhatsAppTue 11:36Action · executed
The ask goes out on the channel they actually answer
Hey — as discussed, let's pull Karim in on Thursday so we can secure the next steps budget-wise.
Written against the blocker, in your words, sent by you. Nothing goes out under your name unapproved.
Follow-upACT-EX-001MR-012EXECUTED - Google CalendarWed 09:12The buyer moves
Karim Bensalah was added to the invite
The buyer did something. That is the only evidence that moves the score — the person who signs is in the room before the meeting, not after it.
SIG-BUY-006EV-004 · strong - NudgeThu 08:00Before the call
A pre-meeting brief is already being written
New attendee, new stakes: what Karim owns, what he is being asked to approve, and the one question worth asking him.
Meeting PrepSG-002 · prep completed - Deal Memory+14 daysThe loop closes
Then it checks whether any of that actually worked
This action asked for an intro, so it is measured on one thing: did the stakeholder join. He did — and the phrasing that worked goes back into Deal Memory.
ProposedAcceptedExecutedMeasuredSucceededstakeholder_expansion_rate ↑EB_intro_rate ↑Buyer MomentumOA-011 · 14-day windowLEARNING_EVENT
Hire one. Add another. Or take the whole team.
Start with the one you hate doing, add the rest when they earn it. Every plan states its monthly capacity. What each one costs is announced on 15 September 2026 — the day it opens.
Hire the one job you hate.
One pipeline story, and it runs without you asking.
For reps carrying hundreds of live opportunities at once.
This is not cold outreach. Every account it opens has something happening on it right now, and a reason you would recognise.
Volume outbound picks everyone who matches the ICP and sends anyway. In the Opener's brief the switch labelled “Cold ICP, no trigger” ships OFF, and stays off unless you turn it on — it will not send into silence on your behalf. Open “how it works” for the reply rates that explain why.
It works before a deal exists, so it is not part of a deal-agent seat. Add it to any plan, including the free Notetaker — the deals it opens land in the same memory the rest of the team reads.
- 01The briefWho counts, and what counts as a reason
Your ICP, your personas, and the list of triggers worth acting on — drafted from the patterns in your own closed-won deals, then edited by you. It is the one thing you own; everything below runs off it.
- 02Intent detectionIt waits for something to happen
A new CRO or RevOps leader inside their first 90 days. A champion who changed jobs and landed somewhere new. A funding round, a hiring spree for AEs, conversation intelligence going in. Each firing trigger is kept with its source, so the reason for the message is auditable rather than asserted.
- 03Two scores, not oneFit is not intent, and intent is not fit
Fit asks whether the account looks like the ones you win. Intent asks whether anything is happening this week. They are scored separately and both have to clear — a perfect-fit account with nothing going on does not get opened, which is exactly the account a volume tool would send to first.
- 04The warm pathHow close you already are
Before it writes anything it looks for the door you already have: degree of connection, mutuals, someone on your side who has worked with them. A champion who moved to a new company is the single strongest opening in the product, because you are not a vendor to that person.
- 05The sequenceWritten per person, not per account
A four-step play across LinkedIn and email — connect with context, the pain note, the proof point, the break-up — paced over nine days, with the angle chosen by what that persona has actually replied to. One prospect is one person: several people at the same account each get their own research and their own message.
- 06You see the repliesDetected, in sequence, replied, booked
It works the top of the funnel on its own and surfaces the ones that answered. When an opening earns a dated, owned next step it graduates into a deal and the rest of the team picks it up — with the trigger, the angle and the thread already in memory.
The reason it refuses to send without a trigger is the bottom row. Same product, same writing, same list — a reason to be there is worth about seven times a well-written cold email.
An active deal is one your buyer touched this month. A reply, an opened doc, a meeting, a commitment. Deals nobody has heard from in 30 days don’t count against your plan — we don’t think they should count in your pipeline either.
Clear limits. Fixed price. Every plan states its monthly capacity above. We’ll tell you before you reach one, and if you need more it’s a block you choose to add — never a surprise invoice.
We also think your forecast is measuring the wrong thing.
That’s a longer argument, and you don’t have to buy it to get value from a single agent. It’s here if you want it — with the arithmetic.
Four minutes of an ordinary forecast call.
One rep, 24 open deals worth $2.37M, a $640k number to hit. Nothing here is unusual and nobody in the room is lying. Watch what the numbers do anyway.
Nobody has ever booked the weighted number.
“Weighted pipeline is $594,750 against a $640,000 number. We’re $45,250 light, but there’s upside in the back half.”
every forecast call, everywhere
That $594,750 is not money anyone is going to send you. It is what this quarter would average if you ran it a thousand times. You get to run it once — and no deal has ever closed at 40%. It closes or it doesn’t.
- $770,000+29% vs the callhit the number
84 of those 1,000 came in within 5% of the call. A typical quarter landed $208,994 away from it, one side or the other. And the thing everyone in that room actually wants to know — do we hit the number? — was 42%. A coin flip. Nobody said “coin flip”.
The gap gets closed in a dropdown.
“Those two are tracking well. I’m moving them to Proposal.”
the same call, four minutes later
Nothing happened with either buyer. Nobody replied, nobody asked for pricing, no second name appeared on the thread. But Qualified is worth 25% and Proposal is worth 40%, so moving one deal a stage adds real money to the company forecast. Try it:
Two dropdowns and the quarter is covered. Nobody committed fraud — the rep genuinely believes those deals are tracking. That is the problem: the only thing the forecast can see is what your reps typed, so the only way to move it is to type something else.
Nobody close-loses a dead deal, and coverage is why.
“We need 3× coverage. We’re at 3.70×. Pipeline’s healthy.”
the Monday pipeline review
Coverage counts every deal that is still open. 6 of these went quiet in April and everyone on the call knows it. Marking them closed-lost would drop the ratio — and the ratio is what gets asked about. So they stay.
The book got worse and both headline numbers got better. And the 3× bar itself only works if you win a third of what you work. This team wins 25.1% of the money in its pipeline, so the honest bar is 3.98×. Hitting 3× here means being under-covered and congratulated for it.
Then the same number spends your marketing budget.
“Channel B sourced $6.00M against Channel A’s $2.00M on the same spend. We’re shifting budget to Channel B.”
the quarterly marketing review
Pipeline sourced counts opportunities created, not deals won. Channel B creates 3× as many. It also closes one in 20, where Channel A closes one in 5.
| Per $60k spent | Channel A | Channel B | Looks better |
|---|---|---|---|
| Opportunities created | 40 | 120 | — |
| Win rate | 22% | 5% | — |
| Pipeline sourced | $2.00M | $6.00M | B, by 3.00× |
| Cost per pipeline dollar | $0.030 | $0.010 | B, by 3.00× |
| Revenue it actually closed | $440k | $300k | A, by $140k |
| Cost per dollar closed | $0.136 | $0.200 | A, by 1.47× |
Channel B looks 3.00× cheaper and closes $140k less for the same $60k. The slide picked the losing channel, and it will pick it again next quarter. Weighting by stage doesn’t rescue it either — everything from both channels enters at Discovery, so both get multiplied by the same 10%.
So stop grading the seller.
Every number in that call came from one place: a stage field your rep owns. Change what the forecast watches and all four problems go with it. The Forecast Agent scores deals on what the buyer did — who replied, who went quiet, who got added to the thread, what was agreed on the call and never happened. It gives you a range instead of a single number, and every Monday it tells you which deals moved, which way, and why it thinks so.
One made-up book of 24 deals, not a customer’s. Every number here is worked out from it in app.jsx, and the 1,000 quarters are fixed — you and the next reader see the same ones.
Don’t pick. Let your pipeline pick.
Drop five open deals in here — any export, any format — and watch the team read them. Each agent takes a pass and reports what it found, and the one with the most to say is the one you should hire first. The file never leaves your browser.
Private beta — nothing to install yet. Nothing to sign up for to read your export. It has its own page at usenudge.app/diagnose if you want to send it to someone.
Show us 5 deals. We’ll tell you.
Drop any export of up to 5 open deals — CSV, Excel, whatever your CRM gave you — and the page reads them back to you. No signup, and nothing to install.
Drop any export here
CSV, Excel (.xlsx), tab-separated, JSON — whatever came out of your CRM. Up to 5 open deals get read.
Pick your CRM and here are the real export steps.
Which agent do you want first?
Two questions. The second one is the useful one: it decides what gets finished first.
The first 100 companies get their last year of lost deals re-read, and a permanent seat in the benchmark they create.