# AI System Inventory Template

**Organization:** [ORGANIZATION NAME]
**Inventory Owner:** [NAME / ROLE]
**Last Updated:** [DATE]
**Version:** 1.0

---

## Purpose

Maintain a single, authoritative catalog of every AI system, tool, model, and embedded AI feature in use across the organization. A complete inventory is the foundation of every AI governance requirement — EU AI Act Article 11, NIST AI RMF "Map" function, and ISO 42001 Clauses 6 & 7 all depend on it.

Use this template to record sanctioned systems, discovered (shadow) systems, vendor-embedded AI, and internally built models in one place.

---

## How to Use

1. Duplicate a row per AI system, feature, or deployment.
2. Complete each column at discovery, and re-review every quarter or when a material change occurs.
3. Flag entries needing legal, security, or privacy review in the **Review Status** column.
4. Feed the inventory into your risk assessment, vendor assessment, and regulatory compliance workflows.

---

## 1. Core Inventory Fields

| # | Field | Description / Example |
|---|-------|-----------------------|
| 1 | System ID | Internal unique identifier (e.g. AI-0042) |
| 2 | System Name | ChatGPT Enterprise, Copilot for M365, custom fraud model v3 |
| 3 | Vendor / Provider | OpenAI, Microsoft, Internal, Third-party API |
| 4 | Category | Generative AI, Predictive, Computer Vision, Embedded AI feature, Agent |
| 5 | Business Owner | Named accountable executive (not a team) |
| 6 | Technical Owner | Engineer / team responsible for ops |
| 7 | Deployment Type | SaaS, On-prem, Hybrid, Edge, API only |
| 8 | Date Introduced | When first used in production |
| 9 | Lifecycle Stage | Pilot, Production, Deprecating, Retired |
| 10 | Replaces / Supersedes | Prior system if any |

---

## 2. Use Case and Purpose

| # | Field | Description / Example |
|---|-------|-----------------------|
| 11 | Primary Use Case | Customer support summarization, code completion, CV screening |
| 12 | Department(s) | Legal, Engineering, HR, Finance |
| 13 | User Population | Number and type of users (employees, contractors, customers) |
| 14 | Business Criticality | Low / Medium / High / Critical |
| 15 | Decision Impact | Informational only / Recommends decisions / Makes automated decisions |
| 16 | Human-in-the-Loop? | Yes / No — describe review step |

---

## 3. Data and Model

| # | Field | Description / Example |
|---|-------|-----------------------|
| 17 | Data Classifications Used | Public / Internal / Confidential / Restricted |
| 18 | Personal Data Involved? | Yes / No — which categories (PII, health, biometric, children) |
| 19 | Data Sources | Systems of record, uploads, third-party APIs |
| 20 | Training Data Source(s) | Vendor data, public data, customer data opt-in |
| 21 | Training Data Opt-out Configured? | Yes / No / N/A |
| 22 | Model / Foundation Model | GPT-4o, Claude Opus 4.6, Gemini 2.5 Pro, internal model v3 |
| 23 | Model Version | Record both model ID and in-use version |
| 24 | Grounding / RAG Sources | Vector stores, connectors, SharePoint sites, knowledge bases |
| 25 | Output Types | Text, code, image, decision score, embedding |

---

## 4. Risk and Compliance

| # | Field | Description / Example |
|---|-------|-----------------------|
| 26 | EU AI Act Classification | Prohibited / High-risk / Limited / Minimal / Not in scope |
| 27 | NIST AI RMF Category | Map / Measure / Manage / Govern evidence link |
| 28 | ISO 42001 Control Reference | e.g. Annex A.6.2.2 |
| 29 | Applicable Regulations | GDPR, HIPAA, SOX, APRA CPS 234, state AI laws |
| 30 | DPIA Completed? | Yes / No / Date |
| 31 | Risk Rating | Low / Medium / High / Critical |
| 32 | Known Limitations | Hallucination rate, bias testing gaps, language coverage |
| 33 | Incident History | Link to incident tickets |

---

## 5. Security and Access

| # | Field | Description / Example |
|---|-------|-----------------------|
| 34 | Authentication Method | SSO, MFA, API key |
| 35 | Access Control | RBAC group(s), role list |
| 36 | Encryption in Transit | TLS 1.2+ confirmed |
| 37 | Encryption at Rest | AES-256 / vendor attestation |
| 38 | Logging Enabled? | Prompt logs, audit logs, retention in days |
| 39 | Data Residency | Country / region of processing and storage |
| 40 | Sub-processors | List per DPA |

---

## 6. Vendor and Contract

| # | Field | Description / Example |
|---|-------|-----------------------|
| 41 | Contract On File? | Yes / No — link |
| 42 | DPA Signed? | Yes / No — date |
| 43 | AI-specific Clauses Present? | Training data use, model change notice, IP, liability |
| 44 | SOC 2 / ISO 27001 / ISO 42001 | Attestation held on file |
| 45 | Renewal Date | For re-assessment trigger |

---

## 7. Review and Monitoring

| # | Field | Description / Example |
|---|-------|-----------------------|
| 46 | Last Reviewed | Date |
| 47 | Next Review Due | Date |
| 48 | Review Status | Pending / Approved / Conditional / Rejected |
| 49 | Monitoring KPI(s) | Accuracy, latency, drift, toxicity, cost |
| 50 | Notes / Open Actions | Free text |

---

## Governance Integration

- Link every high or critical rated entry to a **risk register** item.
- Any change in **Model**, **Deployment Type**, or **Data Classifications Used** requires re-assessment.
- Entries not reviewed in the past 180 days must be flagged to the AI governance committee.

---

## Appendix — Discovery Sources

Populate the inventory from all available sources:

- Identity provider logs (Okta, Azure AD, Google Workspace)
- Network / DNS logs and CASB data
- SaaS spend / procurement records
- Browser extension and endpoint agent data
- Developer environment scans (GitHub Copilot, IDE plugins)
- Employee self-disclosure form (link to `/resources/templates/ai-tool-approval-request`)

---

*This template is provided free by Aona AI. Customize to your organization. See `/resources/templates` for related policy, risk, and vendor assessment templates.*