Loom AI Labs

live Adaptive experience

Behavior Signals

An explainable churn-signal prototype that keeps scoring inputs and intervention rules inside the application.

Behavior Signals live demo interface
Live application capture · verified 2026-07-25
Status
live
Last verified
2026-07-25
AI Fabric
0.4.0
Data
Synthetic

Question

Hypothesis

Can AI explain application-owned behavioral signals without becoming the source of truth for risk scoring?

What this demonstrates

  • Signal values remain visible alongside generated explanations.
  • Application policy constrains the available intervention actions.

What this does not demonstrate

  • Predictive model accuracy
  • Real customer retention outcomes

Architecture and data flow

  1. 01 Synthetic source data
  2. 02 Host application policy
  3. 03 AI Fabric capability
  4. 04 Observable UI result

Guided steps

  1. 01

    Select a synthetic customer profile.

  2. 02

    Inspect the application-owned signals.

  3. 03

    Generate an explanation and compare it with the source values.

Observable evidence

The running application and its health surface are the public evidence for this experiment. The source repository exposes the implementation path behind the visible result.

Security and data boundary

The host calculates behavioral signals and defines allowed interventions. AI Fabric structures explanations and proposed next steps.

Customer and engagement records are synthetic.

Known limitations

  • The signal model is illustrative, not statistically validated.
  • No production customer data is processed.

Technologies and products used

  • Spring Boot
  • AI Fabric 0.4.0
  • Structured generation
  • OpenAI

Connect

Inspect the live system and its source together.

The experiment is useful when the visible result, application boundary and implementation can all be questioned.