Cold email written like the top 1%, because it learned only from them

A sibling model to ViralML™, fine-tuned exclusively on the top 1% best-performing cold emails by positive reply and close rate. Same Frankfurt datacenter. Same obsession with outliers.

Jonas
H
Sequence · Q4 RevOps leaders
Predicted reply 18.4%
To
Subject
Step 1 of 469 wordsReading level: grade 7ColdLLM-1.0 · fra-1
Top 0%
of cold emails by positive reply rate and closed-won: the only emails in the training set.
2.9M emails reviewed · 29k kept
+0%
higher close rate for Halden Systems in the first full quarter.
see case study ↓
0×
positive reply rate vs. clients' previous hand-written sequences (median).
across 14 pilot clients
0
spam-trigger phrases. Deliverability checks run before a single send.
EU-hosted · GDPR-native

From a list of names to a sequence worth replying to

Five steps between your ICP and a first email that reads like it was written for one person, because it was.

01
Tell it who and what

Your ICP, your offer, three proof points. That's the whole setup.

02
It finds the trigger

Hires, funding, job posts, tech-stack changes: public signals that make 'why now' obvious.

03
It writes like the top 1%

Trigger, quantified pain, one peer proof, a low-friction ask. Under 90 words. Every time.

04
Pre-flight before a single send

Spam triggers, merge-tag leaks, link count and reading level: checked, fixed, logged.

05
It learns from your replies

Positive replies feed back into your workspace's variants. Your sequence gets sharper every week.

Paste your worst cold email. We'll be gentle.

Pick an example or paste your own. We score it live, flag what the top 1% never do, and show you the rewrite with a predicted reply rate.

H
Rewrite
Write for
Draft
Positive reply 1.1%
Offer: Halden · RevOps platform · rewrites are pre-generated in this demo; the score on your draft is live.
ColdLLM™ rewriteColdLLM-1.0 · fra-1

Hit Rewrite to see the top-1% version.

Positive reply
1.1%→…
Reply
3.8%→…
Words
72→…
Grade
11→…
Issues
6→…
Trigger eventOpens on something that happened to them, not on you. Top-1% emails do this 4.2× more often.
Quantified painA specific number they already feel. Vague pain gets archived.
Peer proofOne named peer, one result. No logo soup.
Low-friction askAn interest-based question with a time-box. 2.3× more positive replies than “book a call”.

A typical first email

Everything between 'who' and 'booked'

Trigger events

Opens on what just happened to them. Top-1% emails do this 4.2× more often.

Four-step sequences

Follow-ups that add a new angle instead of 'just bumping this'. Step 3 is usually the one that books.

Persona variants

One offer, rewritten for a CISO, a RevOps lead and a founder. Different pain, different proof, different length.

Deliverability guard

Spam triggers, link count, merge-tag leaks and reading level, all checked before anything leaves.

CRM & sequencer export

HubSpot and Salesforce sync. CSV for everything else. Or call it from the API.

Reply coach

When they answer, it drafts the reply that gets the meeting, not the one that gets 'not right now'.

How Halden Systems closed 24% more deals from the same pipeline

H
Halden Systems
B2B SaaS · RevOps software · Munich · 140 employees

Halden's 9-person outbound team was booking plenty of meetings, but too many of them were the wrong meetings. Prospects showed up curious, not in pain. Deals stalled at evaluation.

In Q2 2026 they moved every outbound sequence to ColdLLM™. Same ICP list. Same reps. Same pricing and offer. The only variable was the email.

“The emails pre-qualified people for us. Prospects showed up already knowing their number. Our AEs stopped doing discovery and started doing closing.”
JW
Jonas Weber
VP Sales, Halden Systems
Close rate · opportunity → closed-won
+0%11.3% → 14.0%
Q1 2026 · hand-writtenQ2 2026 · ColdLLM™
Positive reply rate
4.1%8.6%
Meeting → opportunity
38%50%
Opportunity → won
11.3%14.0% · +24%
Sales cycle
71 days55 days
€1.3M
added closed-won ARR in one quarter
−38%
rep time spent writing sequences

Same prospect. Same offer. Different email.

Before · hand-writtenSubject: Halden x {{company}} : quick intro

Hi {{first_name}},

I hope this email finds you well! My name is Tobias and I'm reaching out from Halden Systems, the leading AI-powered RevOps platform trusted by innovative companies across Europe.

We help revenue teams streamline their operations, improve forecasting accuracy, and unlock growth through data-driven insights and seamless integrations.

Would you be open to a 30-minute call next week to explore how we could help {{company}} achieve its goals?

Best,
Tobias

Reply 3.8%Positive 1.1%Words 72
After · ColdLLM™Subject: your forecast call on thursdays

Hi Clara,

Noticed Veltra just hired its third RevOps analyst this year, which usually means the Thursday forecast call is taking half a day to prep.

Kinetik had the same problem at your size. They cut forecast prep from 6 hours to 40 minutes, and their commit accuracy went from 71% to 93%.

Worth a 12-minute look at how? If not, no worries, happy to send the one-pager instead.

Tobias

Reply 14.9%Positive 8.6%Words 70

More from the pilot

Kinetik · Logistics SaaS · Berlin
3.4×
meetings booked per 1,000 sends

Replaced a 7-step generic sequence with a 4-step ColdLLM™ sequence triggered by warehouse job postings.

Fernweh · Travel tech · Hamburg
−46%
cost per qualified opportunity

Fewer sends, better targeting, and emails that read like a human wrote them at 11pm, because a human would.

Orbital · Cybersecurity · Zürich
+31%
positive replies from C-level

Executive-tier variants trained on the top 1% of emails that got CISOs to answer. Shorter. Blunter. Better.

Specs

Training set
2.9M cold emails reviewed · 29k kept · top 1% by positive reply and closed-won
Model
ColdLLM-1.0 · sibling of ViralML-1.0
Languages
English, German
Sequences
1–6 steps
Signals
Public hiring, funding, news, job posts, tech stack
Pre-flight
Spam triggers, links, merge tags, length, reading level
Integrations
HubSpot, Salesforce, CSV, API
Compliance
GDPR-native · opt-out line on every step · fra-1 hosting

FAQ

Is cold email legal in the EU?+

B2B cold email is generally possible under legitimate interest in most member states, with a clear opt-out. ColdLLM™ adds one to every step. This isn't legal advice, so check with your DPO.

Does it send the emails?+

No. It writes and scores; you send through your existing tools. That keeps your domains, your warm-up and your data where they are.

Will prospects know it's AI?+

They'll know it's short, specific and about them. That's what the top 1% have in common.

Do you train on my emails?+

Not by default. You can opt into a private fine-tune that stays in your workspace.

What reply rate should I expect?+

Across 14 pilots, a median 2.1× your previous positive reply rate. List quality still matters more than anything we write.

More from ViralML

Your competitors already get cold emails. Send better ones.

You're on the list. We'll be in touch from Frankfurt.

Brands already signed up for early access