# Private AI Deployment Decision Matrix

Use this to choose a deployment pattern by data sensitivity and operational capacity.

## Data Classification

| Data class | Examples | Default AI boundary |
| --- | --- | --- |
| Public | Published pages, public docs | Approved SaaS |
| Internal | Internal notes, anonymized examples | Enterprise SaaS |
| Confidential | Customer data, contracts, source code, financials | Enterprise SaaS with controls / VPC / self-hosted |
| Restricted | Health, legal privilege, HR investigations, credentials, regulated records | Legal/security review; often local/VPC/no AI |

## Pattern Comparison

| Pattern | Privacy | Capability | Cost | Ops burden | Best fit |
| --- | --- | --- | --- | --- | --- |
| Consumer SaaS | Low | High | Low | Low | Public/personal work |
| Enterprise SaaS | Medium | High | Medium | Low | Normal company work |
| VPC/private cloud | High | Medium/high | High | Medium | Confidential workflows |
| Self-hosted inference | Very high | Medium/variable | Variable | High | Restricted/custom/scale |
| Local-device model | Very high | Lower | Low/medium | Low/medium | Narrow sensitive tasks |

## Decision Questions

- What data enters the model?
- What output impact exists?
- What quality is required?
- What latency is required?
- Who operates it?
- What proof do customers or regulators need?
- What is the fallback if the private model is not good enough?

## Routing Rule

| Use case | Data | Output impact | Chosen boundary | Owner |
| --- | --- | --- | --- | --- |
| | | Draft / recommendation / action / decision | | |