# Practical Enterprise AI Governance Framework

## 1. Value and use-case tier
- What decision or workflow improves?
- What measurable value is expected?
- What is the impact if the AI output is wrong?

## 2. Information boundary
- Which sources are approved?
- Is confidential, personal, regulated, or production data involved?
- What must never enter the model or knowledge base?

## 3. Accountability
- Who owns the use case, model behavior, data, decision, and operation?
- Where is human review mandatory?
- Who may override or stop the workflow?

## 4. Evaluation
- Define accuracy, completeness, citation, false-positive, and false-negative expectations.
- Test representative, boundary, adversarial, and low-information cases.
- Record human overrides and reasons.

## 5. Operation
- Monitor quality, value, adoption, latency, cost, incidents, and drift.
- Maintain version, change, and retirement controls.
- Review controls when the use case, data, model, or consequence changes.

Principle: AI recommends. Accountable leaders decide. Teams deliver.