Long-Term Alignment for AI

​

Series A

AI systems are incredible at delivering on timely, high-volume training signals

Training AI on signals that yield more than a “sugar-rush” signals is a $100B opportunity

Churney: Long-term signal alignment for AI

AI systems are optimized towards sub-optimal short-term training signals

​

Consume Data warehouses

Send predictive signals

Control and perfect the outcome

30% lift in Long-Term ROI

for our customers

Commercial partnerships

Solution

The outcome

Problem

“what you're building is exactly what the industry needs”

Jerry Dischler, ex-head of Google Ads

Churney

Series A Deck

1. Search. Ad platforms. LLMs.
2. Different AI, same destination: End-to-end training for real-world outcomes
3. The bottleneck isn't the model, It's the training signal

CPC

tROAS

Long-Term Optimized AI

CPA

Pre-training

RLHF

RLVF/RFT

AI systems converge

Churney

Series A Deck

We have done it before

1. Ad platforms: big market, fast experiments, big upside
2. Conversion APIs mature, Data warehouse access easy
3. Mature, automated system with one big lever: �Communicating long-term value for each advertiser
4. Churney’s feedback loop generates $2m a month in incremental revenue for our clients

3.88x

YoY growth

Financial Results

~2.2m

Monthly incremental revenue

B2B

Subscription apps

ecommerce

gaming

Churney

Series A Deck

Churney’s world view

Same structural gap, Same solution

1. Long-term value based fine-tuning is inevitable
2. LLM builders get reward-hacked when fine-tuning
3. Companies using LLMs will suffer

Fine-tuning towards long-term revenue requires specialized signal infrastructure.

• Churney solved this for ads AI, and owns the data and feedback loops missing for LLMs

Real ChatGPT response after OpenAI optimized for short-term user feedback (thumbs-up/down). Rolled back in 4 days.

Reward hacking->misalignment problem

Churney

Series A Deck

The vision

Propelling all AI systems to deliver long term goals

The signal layer: where the real leverage is

Post-training is only as good as the data it learns from. Not all data is equal.

COMPANY

WHAT THEY PROVIDE

HOW THEY GET IT

SIGNAL QUALITY

Scale AI

$14.3B Meta deal

Human-labeled training data at scale

Crowd-sourced annotators, managed labeling ops

Generic labels. No causal link to business outcomes.

Surge AI

~$1B raising

Premium RLHF preference data

Expert human raters ranking model outputs

High-quality preferences. Still short-term, thumbs-up/down signals.

Mercor

$350M Series C

Domain expert feedback for fine-tuning

30K+ specialists (doctors, lawyers, scientists)

Deep expertise, but point-in-time. No outcome measurement.

Churney

Series A

Causally robust long-term outcome signals

Live production experiments on real user behavior

Measures what actually drives long-term value. Proven on ad platforms.

The gap: Everyone else provides human opinions about what's good. Churney provides causal predictors of what actually works — the same signal infrastructure we built for $650B in ad spend.

5% of long-term focused performance marketing (~$650B) and Gen AI (~200B by 2027 growing rapidly) spend

Reachable AI systems in e-com and mobile-app verticals across US, EU, and key growth markets.

Near-term obtainable share based on existing pilots, conversion rates, and partner reach (Meta, Google, TikTok, GCP), LLM projects with existing clients

SOM

~$150M

SAM

~$10.0B

TAM

~$42B

Total Opportunity

Churney

Series A Deck

Founders

Past, Present, Future

Brian Brost, PhD;

Co-founder, CPTO

Spotify Research, Causal ML

Cracked causal feedback loops in predictive AI training

Define product and R&D vision

Noy Rotbart, PhD;

Co-founder, CEO

Liveintent, built Supply-Side Platform

Sold Churney to multiple verticals/ platforms

GTM, Partnerships

World class research team in causal AI/ML with 10 PhDs

Churney

Series A Deck

Gert Lanckriet

Head of AI

​

Mehdi Ghissassi

x-Head of Product

Oliver Koch

PM Director

Kieran Flanagan

Marketing AI SVP

Angels

Funds

Nacho Floristan

Director of GenAI,

Joonas Hjelt

Director, superintelligence

Lasse Espeholt

Eng lead

Backing

Churney

Series A Deck

Already existing client - Hubx Nova chatbot app

Goal: Maximize free trial conversions

Problem: Base Gemini models don't know what drives conversions

Churney Reinforcement Fine-Tuning Signal

Score millions of (prompt, response) pairs based on predicted impact on a user's likelihood to convert.

Result

Fine-tuned Gemini that drives measurably improved trial conversions

GenAI Signal - First example

2025

pLTV optimized signal

2026

2027

$60M

+$300m

​

+$1b

​

Series A - $15m to maximize aligned signal

Churney monetization is currently ~5% of optimized signal.

Churney

Series A Deck

Appendix

Churney

Series A Deck

Churney

Series A Deck

Churney

Series A Deck

Churney

Series A Deck

Churney

Series A Deck

Churney

Series A Deck

Zapier: sticky enterprise adoption

​

$96k in ARR

Leadtech: 5.8x ARR growth in 5 months

​

$696k in ARR across 5 apps

$156k in ARR

​

2 apps onboarded

HubX: immediate post-pilot upsell

​

$156k in ARR across 2 apps

Signal land and expand

Churney

Series A Deck

Tie our revenue to the amount of signal used (Currently 2.5%)

Deep Platform Partnerships

• Joint incentive for AI makers to make Churney a success
• Led by data scientists, adopted by commercial, influencing product
• Partners fully financing prospect’s pilots, easy to replicate
• Gemini’s first pilot - a 700k grant to test GenAI signal engineering

​

GTM

Churney

Series A Deck

Solution architecture

3 years of live experiments on +$100M ad spend as a validation of the hypothesis

Churney

Series A Deck

Gaming

E-comm / Retail

Subscription Apps

SaaS

Consumer

Clients & Pilots across verticals, globally

Churney

Series A Deck