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

# What Helix Does

> How Helix connects enterprise data to an intended AI use without taking over customer authority.

Helix helps enterprise teams qualify a bounded collection of content for one
declared AI use, then preserve what they decided and why.

Sensitive Data Intelligence is the semantic foundation. It produces attributable
evidence from unstructured content, but it is not the entire product and does
not replace customer policy.

## A deliberate authority boundary

Helix and the customer have different jobs:

1. **Helix produces attributable evidence** and accounts for unsupported,
   partial, failed, and unresolved cases.
2. **The customer defines the intended use and policy**, evaluates the evidence,
   and makes the decision.
3. **The customer controls downstream use** while Helix preserves the approved
   population and the basis for the decision.

This separation matters. A bank account number can be identified without
claiming that every appearance violates policy. Sensitivity, permitted use, and
required action depend on context and the customer's governing rules.

## Evidence, not forced certainty

A credible system must represent what it could not establish. Helix preserves
abstention, unsupported content, unavailable processing, and failed processing
instead of silently converting them into clean results.

<Info>
  A classification describes what the evidence supports. It does not decide
  whether content is prohibited, permitted, or ready for a particular use.
</Info>

## Continue with the model

Read [Corpus Qualification](/concepts/qualification) for the decision loop,
[Evidence and Uncertainty](/concepts/evidence) for the evidence model, and
[Classification Dimensions](/concepts/classification) for the semantic tasks
that support sensitive-content understanding.