> ## Documentation Index
> Fetch the complete documentation index at: https://docs.manthan.systems/llms.txt
> Use this file to discover all available pages before exploring further.

# From Signals to Authorization

> How AI-generated signals are transformed into deterministic authorization decisions and enforced execution

This document explains how Parmana transforms AI-generated signals into deterministic Authorization Decisions that control real-world execution.

***

## System overview

Parmana is a deterministic authority verification system.

It ensures that AI-generated signals cannot directly trigger execution without governance evaluation.

***

## System flow

```text theme={null}
AI → Signals → Governance → Authorization Decision → Execution Runtime → Attestation
1. Signal generation

AI systems produce raw outputs such as:

recommendations
predictions
proposed actions

These outputs are not trusted or executable.

2. Signal processing

Raw AI outputs are transformed into structured signals:

normalized
schema validated
provenance tagged
integrity checked

Only structured signals proceed forward.

3. Governance evaluation

Governance receives Verified Signals and evaluates them against:

signed policies
deterministic rules
authorization constraints

Governance does NOT use probabilistic AI logic.

4. Authorization Decision

Governance produces a deterministic decision:

APPROVE
REJECT
REQUIRE_OVERRIDE

This decision is fully reproducible.

5. Execution Runtime enforcement

Execution Runtime enforces the decision:

validates executionId
enforces replay protection
ensures decision integrity
blocks unauthorized execution
6. Attestation layer

Every decision produces a cryptographic attestation containing:

policy version
signal hash
decision outcome
runtime identity
executionId

This enables independent verification.

Core insight

The system enforces strict separation:

AI generates signals
Governance evaluates signals
Execution Runtime enforces decisions
Why this matters

Without this separation:

AI could directly influence execution
decisions would be non-deterministic
auditability would break
authority would be unclear
Deterministic guarantee

The same inputs always produce the same output:

same signals + same policy → same Authorization Decision
Summary

Parmana ensures:

AI is isolated to signal generation
Governance determines authority
Execution is strictly enforced
Every action is verifiable
```