JB
Jeff Brokaw // Lab
THE OUTBOUND MACHINE
NAMES REDACTED · REAL MACHINE
← back to lab
// The Outbound Machine

Nobody reads your deck. Be honest, you don't read theirs either.

So this doesn't send decks. Point it at a market and, for every named account, it reads the public signals (funding, hiring, the founder's posts, the status page), reasons out the gap, builds them a working proof of it in their own brand, mails it to their desk with a unique QR Code, then reacts to what they do with it. No invented data. No dashboard access. Just what anyone can see, used well.

It cost $1.12 to put there. Your last click cost $14 and got ignored.
If you don't read flowcharts, here's the whole machine in one line.
  1. 1Research a named company from public signals
  2. 2Score the fit and kill the bad ones before spending a cent
  3. 3Build them a working proof in their own brand
  4. 4Mail it to their desk with a unique QR Code
  5. 5React automatically the moment they engage
01

The machine, as it actually runs

the deep version · click any node to inspect
the-outbound-killer · 38 nodes · reacts to behavior
+
Parameters
Settings
Input / Output
Sources Clay · enrich + score Opus 4.8 · research n8n builds the proof Deliver Decision tree · HubSpot
0
SOURCED
target accounts
0
KILLED AT GATE
no spend
0
PROOFS BUILT
microsite + mail
0
DELIVERED
desk + LinkedIn
0
ENGAGED
scanned, deal moved
0
meeting-ready
Per proof on a desk: $1.12 · the click they ignored: $14
$0
to run the whole campaign · 12 targets, 7 refused at $0 spend
$0
cost per meeting-ready account · vs ~$300+ typical B2B SDR cost per booked meeting
StageEventTool
Kill gate 7 killed · 5 pass · no spend below fit 70 Clay
Research Dossier assembled from public signals · high confidence Opus 4.8
Proof built Postcard + white microsite deployed · QR registered Stannp · $1.12
Delivered QR scanned · 4m 20s on page · threshold tripped GA4 · Stannp
Decision Deal → Hot · owner alerted · LinkedIn queued HubSpot · Slack
02 First it researches, from public signals only press Run to watch it build

The kill gate

Clay · fit ≥ 70 · no spend below

The dossier, reasoned from what anyone can see

Opus 4.8 · every fact sourced
ACCOUNT
Prospect Co · AI agent infrastructure
STAGE
Series A · ~40 people
PUBLIC SIGNALS (nothing here needs their dashboard)
$22M Series A, announced this springfunding wire
6 open backend / infra roles, 0 in SRE, reliability, or observabilityjob boards
Founder posts about agent volume and shipping speed, never uptime or failuresLinkedIn
Public status page logged 3 incidents in 90 daysstatus page
Job posts list a sprawling stack, no observability tooling namedtech stack (JDs)
THE INFERENCE high confidence
Shipping agents into production faster than they can see them fail.
Scaling compute 6-to-0 over reliability, incidents already public, no observability anywhere in the stack. A team this size, moving this fast, is flying blind on the one thing that breaks trust: what happens after deploy. The gap is visibility, and it compounds with every service they add. Every line above traces to a public signal, never an invented internal number.
03 Then it builds the proof: physical meets digital postcard lands on the desk · the QR opens the page

The postcard on the desk

Stannp · dynamic QR · ~$1/lead
copy kill gate · machine-built, human-grade
REJECTEDHi there, hope you're well! We help AI teams with observability. Would love to connect.
rejected: generic, indistinguishable, AI slop
SHIPPEDYou posted 6 backend roles and 0 observability hires this quarter. Here's what that's about to cost you.
FOR THE FOUNDER · PROSPECT CO
You posted 6 backend roles and zero observability hires. Here's what that costs.
the vendor · agent observability

We show you every agent failure before your users do.

Scan it. We modeled the cost of flying blind on your own public numbers.

unignorable.jeffbrokaw.com/p/redacted
per-recipient QR
This proof cost $1.12 to land on a desk.
Your last click cost $14 and got ignored.
~$1.00 print + mail (Stannp) · $0.12 compute. Honest dollar-for-dollar, no invented lift.
scan
opens
unignorable.jeffbrokaw.com/p/redacted
Built and deployed by the workflow, in the prospect's own brand generated just now
prepared for the founder
You're shipping agents faster than you can see them fail.
Everything below is reasoned from your public footprint. No dashboard access, no guesswork, just what anyone can see if they actually look.
$22M
Series A, scaling fast
6 : 0
backend vs reliability roles open
3
incidents on your status page, 90 days
The stakes: your peers already see this
A direct competitor publishes a 99.9% uptime SLA and a live status page, and leads with reliability in its pitch.source: their status page + marketing site
Another peer just opened a Head of Reliability role.source: their public job post
Every incident your users hit, your competitors get to say they don't.
What flying blind costs, modeled from your public signals
Modeled cost / month
$16k/mo
~108 engineer-hours a month spent finding failures by hand
Modeled from public signals (your headcount, your status page's incident cadence, market-rate detection time without observability) at a blended $150/hr. Your real numbers differ; the shape does not. The faster you scale, the steeper this climbs.
How we know, signal by signal
Hiring 6:0 Status page Founder's posts
A 30-second note, for the founder
0:28
Production ships a 30-second note from the vendor's founder via ElevenLabs.
This proof cost $1.12 to land on your desk. Your last click cost $14 and got ignored.
Built for the prospect by the vendor, agent observability. Both names redacted.
Clean white, the prospect's own logo, palette, and font, generated on the fly by the workflow. Deliberately nothing like this dark Lab site, because the page is the recipient's, not the sender's.
04 Then it reacts to what they do Stannp scan + GA4 → HubSpot → the owner

The scan fires, and the machine reacts

scan → HubSpot → the owner
HUBSPOT DEAL
Proof Sent
ON SCAN
Engaged
DEEP ENGAGEMENT
Hot
ON REPLY
Conversation
#gtm-alerts · deal ownerjust now
The prospect's founder scanned 2 minutes ago. 4m 20s on the page, ran the cost-of-blindness model out to 40 services, read the reliability comparison twice. This is the moment.
SUGGESTED NEXT ACTION
Call now. Drafted, behavior-aware opener attached below.
HubSpot: Proof Sent → Hot task created · due now PURL + full trace on the deal
THE TREE, RESOLVED FOR THIS ACCOUNT (taken path lit, alternates armed)
scanned in 7 days?yes · deal → Engaged, owner alerted, task createdTAKEN
deep engagement?yes · deal → Hot, LinkedIn follow-up queued, times offeredTAKEN
shallow scan only?deal stays Engaged, lighter nudge, second value touch
not scanned in 7 days?one considered second touch (new postcard / LinkedIn opener)
they reply?Opus classifies the intent, then routes four ways
interesteddeal → Conversation, offer times
objectionrebuttal drafted from the dossier, queued to the AE
referralre-target the named person with a fresh proof
not nowdeal → Dormant, dated re-engage on a fresh signal
DRAFTED FOLLOW-UP · BEHAVIOR-AWARE · QUEUED TO THE AE (MANUAL SEND)You ran the cost-of-blindness model out past 40 services and read the reliability comparison twice. That gap, the one your peers already publish around, is the conversation I'd want to have. Fifteen minutes this week to show you what we catch?

The LinkedIn note

MANUAL SEND
inDraft note → the founderqueued to the AE
I pulled your public footprint last week. $22M Series A, 6 backend roles open and 0 in SRE, and your status page logged 3 incidents in 90 days. You're shipping agents faster than you can see them fail. I didn't write you a pitch. I built you something: unignorable.jeffbrokaw.com/p/redacted It models the cost of flying blind on your own public numbers, and shows what we catch before your users do. Three minutes.
No API send. A human reads, edits, and sends it by hand. That gate is in the architecture, on purpose.

This is a demo of a machine that is real.

Both names are redacted here, the way you'd protect a live client, because the machine is real and runs against real targets. The figures are illustrative, but every fact the machine surfaces is the kind anyone could pull from public signals: a funding wire, a job board, a status page, a founder's posts. Nothing claims a private metric the stack could never see. The pipeline is not a mockup either, and it is sender-agnostic: point it at a market and any vendor's signals and offer slot in. Run it and the "could you do this for us?" answer is an honest yes.

The line, end to end
n8n orchestrates the whole thing./Clay enriches, scores, and kills the weak fits before a cent is spent./Opus 4.8 reasons the dossier from public signals and later classifies the replies./the workflow builds and deploys the page, voiced by ElevenLabs./Stannp mails the postcard./HubSpot moves the deal and pings a human only when it matters./GA4 watches the page.