Step 1
Open the workspace
Start in the Fraud workspace and confirm the project-backed flow for scenario, configure, score, and results.
Operational AI Quickstart
Fraud / Risk Scoring visual quickstart
Choose a Fraud project, stage training data, train or activate a saved version, score new rows, and review risk outputs with provenance.
Visual walkthrough
Follow the workflow step by step, using the screenshots where available and the concise action notes throughout.
Before you start
Select or create the right Fraud project first.
Plan to train or activate a saved version before the scoring step.
Step 2
Stage the training scenario
Choose the project and upload a labeled training dataset or starter sample.
Step 3
Train and manage versions
Use Configure to train a saved model version, review lifecycle evidence, and set activation state.
Step 4
Score new rows
Stage the scoring dataset and run scoring against the active model or a selected saved version.
Step 5
Review risk outputs
Inspect row-level probabilities, decision context, and grounded AI interpretation in Results.
Expected outputs
Row-level risk scores and decision context
Saved-model provenance such as version and feature schema
Result summary and explanation surfaces for the scored run
Common mistakes
Confusing training and scoring dataset roles
Trying to score before a ready saved version exists
Reviewing results after changing execution source without rerunning
Next steps
Move from walkthrough to the right working page
Open the live workspace when you are ready to execute, or use the related product, use case, and spec pages when you need more context first.
Navigate
Keep this quickstart connected to the resource library
Return to the quickstart index for another product walkthrough, or go back to the resource hub for use cases, input specs, and how-to guides.