AI Glossary Term

What does “Shadow Deployment” mean?

Definition

Running a model in parallel with production traffic without affecting user-facing decisions.

Browse the full AI Glossary Explore all free AI guides

How Shadow Deployment works

Example

A bank runs a new fraud model alongside the current one on real transactions. The new model's decisions are recorded but not acted on, and the team compares them before switching.

Why it matters

It tests a model on real traffic without exposing users to its mistakes. It costs extra compute while running.

Common questions

Shadow vs A/B test?

In a shadow test, the new model's output is not shown to users. In an A/B test, some users receive it.

What does it reveal?

How the new model behaves on real inputs, including rare cases that test data lacks.

What are its limits?

Users do not react to the new output, so effects on behavior cannot be measured.

Speculative DecodingAn inference acceleration method where a small draft model proposes tokens that a larger model verifies in parallel.SentencePieceA language-agnostic tokenizer that learns subword units directly from raw text without pre-splitting on whitespace.Autonomous SystemA system that can make decisions and act with limited or no direct human control in real time.Continual LearningTraining approaches that let a model keep learning from new data without forgetting prior knowledge.Intent ClassificationPredicting the user's purpose from a text query to route it correctly.Recommendation SystemA model pipeline that predicts user preferences for ranking content or products.

Learn more in our free guides

Canary and Shadow Deployments Shadow AI in the Workplace Model Validation Gates Before Deployment Blue-Green Deployment for ML Models

Test yourself

Canary and Shadow Deployments Quiz Shadow AI in the Workplace Quiz Model Validation Gates Before Deployment Quiz