✈️Co-pilot
Co-Pilot is our proprietary AI engine that automatically predicts and applies the best retention, upsell, and engagement strategies for each unique customer profile.
What is Co-Pilot?
Co-Pilot by Churned is an automated learning engine that acts as the "brain" inside your existing tech stack. It’s designed to enhance your tools, not replace them, by automatically finding and deploying the next best action for every customer.
Learning from Your Data
Co-Pilot builds a predictive model by combining all your existing behavioral, transactional, and demographic data to learn what actions (messages, offers, timings) truly work best for each specific customer segment.
It connects to:
CRMs and automation tools such as HubSpot, Klaviyo, and Customer.io
Data warehouses and CDPs
Analytics and event tracking systems
Deciding the Next Best Action
You maintain full control by providing Co-Pilot with a "menu" of possible actions it can take within your workflows.
Once you define the actions, Co-Pilot then decides:
Which action to deploy
When to deploy it
Through which channel
Triggering Actions in Your Stack
The model automatically sends the next best action back to your marketing tools to be executed. It pushes these decisions directly into your existing workflows, triggering the right action for the right customer without manual intervention.
For example, a "post churn" flow might have three options you've defined:
Action A: Send 15% discount email.
Action B: Send a personalized win back message.
Action C: Wait 3 days and send a survey.
Applying Across the Customer Journey
Co-Pilot is designed to learn, act, and optimize across every key stage of the customer journey.
Co-Pilot can be used for:
Onboarding: Personalize early interactions and improve first week engagement.
Engagement: Find the best timing, channel, and content for continued activity.
Reactivation: Test and automate the best win back strategies.
Retention: Continuously refine offers and communication to prevent churn.
How It Works: The Two Phase Learning Engine
Co-Pilot operates in two dynamic phases — Exploration and Activation — that together form a self learning engagement engine.
1. Exploration: Learning What Works
In this phase, Co-Pilot strategically deploys multiple actions (like different discounts, win back emails, or in app nudges) across diverse customer segments.Think of it as an automated, multi variant test.
By testing multiple strategies simultaneously, Co-Pilot collects a wide range of unbiased data to power its predictive model and avoids the limitations of traditional A/B testing.
It automatically:
Tests multiple interventions (emails, discounts, nudges, etc.) across diverse customer segments.
Collects unbiased behavioral data to train an initial predictive model.
Identifies what works best for each type of customer, not just the average one.
2. Activation: Acting on the Best Strategy
Once the model has learned what works, Co-Pilot enters the activation phase, automatically applying the best next action for each customer segment.
When a customer reaches that point, Co-Pilot analyzes their profile and history to determine the "next-best-action" with the highest probability of success.
It automatically:
Selects the optimal treatment per customer (e.g., “send WhatsApp message” vs. “offer discount”).
Pushes that action directly to your connected marketing tools.
Continuously retrains on new outcomes to adapt to changing behaviors and seasonality.
Continuous Exploration (The Built in Control Group)
Co-Pilot maintains a persistent exploration group, meaning a small, predefined portion of your audience (e.g., 5%) is always reserved for continuous testing, even after an optimal strategy is identified. This function serves as a dynamic control group, providing a constant stream of new, unbiased data for model retraining.
This ensures the model can mitigate drift, adapt to evolving customer behaviors or seasonality, and continuously validate its own strategies, guaranteeing that your campaigns remain dynamically optimized against the most current, real-world responses.
Getting started with Co-Pilot
Setting up Co-Pilot is a simple, five step process:
Connect your data sources: Link your behavioral, demographic, engagement, and purchase data.
Define your action options: Specify the actions Co-Pilot is allowed to test (e.g., channels, offers, message types).
Deploy your first exploration phase: Co-Pilot starts experimenting immediately and learning from real time interactions.
Move to activation: As the model gains confidence, insights are applied automatically across your campaigns.
Monitor results: Watch how Co-Pilot adapts and automatically improves your key retention metrics.
Last updated
Was this helpful?