Fraud and riskLive

Flow draft generation

An assistant that turns the policy text you write into a fraud flow draft.

Generates a starter flow draft from the policy text you write in the panel. The draft isn't saved anywhere, starts out disabled and isn't published; to go live, you must save it and take it through validation, simulation and the publishing gate.

Model card

Automatic publishing
None
Side-effect nodes
Never generated
Validation
Mandatory
Personal data
Not sent
Capabilities

What does it do?

  • Draft from text

    A starter flow from your policy text.

  • Starts disabled

    The draft isn't saved or published.

  • No side-effect nodes

    Cannot generate nodes that run code, make external calls, query databases or emit events.

  • Engine validator

    A draft that fails the validator is never returned.

  • Audit stamp

    The model and prompt version used are carried on the draft.

  • No personal data

    Only the policy text is sent to the model service.

Pipeline

How does it work, step by step?

  1. 1

    Stage 1

    Policy text

    The operator writes the rule in the panel.

  2. 2

    Stage 2

    Model

    The text goes to the model as untrusted input.

  3. 3

    Stage 3

    Validator

    The draft passes through the engine's validator.

  4. 4

    Stage 4

    Disabled draft

    The draft is returned as the response.

  5. 5

    Stage 5

    Human

    Saving, simulation and publishing stay with the operator.

Meet Flow draft generation.

Let's review with our team how the model would work in your business.

Flow draft generation — AI | Lunixi