AI model inventory and approval workflows

What vendors state their systems record and how approvals, audit trails, and exports work. Plan inclusion and live tenant behavior require demo verification.

ModelOp: follow the owner, approval and export

ModelOp’s public manuals fill a previous coverage gap. These are documented capabilities, with plan details still to confirm.

ModelOp

Ownership and fields
Use-case records include Owning Group, Risk Tier and Organization. Administrators can add company-specific forms. SourceThe bank still chooses accountable people and required fields.
Approval trace
Approval records show a name, status and links to the approval item, typically in Jira or ServiceNow. SourceExternal ticket links do not prove the linked system is included in a quoted plan.
Documentation
Generate a model card from the use case’s known details, test results and metadata, manually or through a lifecycle workflow. SourceA generated card does not independently validate the model or establish regulatory approval.
Risk rules and review cadence
Optional utilities define risk-tier rules and create controls or tasks from model metadata, such as quarterly versus annual reviews. SourceThe documentation labels these utilities optional; confirm licensing and configuration.
Inventory export
Download the current use-case or implementation list as CSV, with or without filters. SourceAn inventory extract is not a complete audit-evidence export.
A fair question

Who owns the work once you buy the software?

The vendors describe roles, approvals and audit trails. That helps you compare the tools. It doesn’t decide who owns your program or show that a template meets your bank’s needs. The software alone won’t tell you whether you’re ready for an audit.

See which frameworks vendors name →

What each platform says you can inventory

VerifyWiseSource record

Vendor FAQ says system records can represent simple ML models, generative systems, and third-party vendor solutions, with documentation, assessments, and risks.

Does not establish SR 26-2 mapping or bank suitability

Credo AISource record

Credo public SDK reference names Models, Use Cases, Vendors and Custom Fields resources/types.

Does not establish hosted-product entitlement or bank configuration

ValidMindSource record

ValidMind platform page describes an enterprise model inventory across statistical/traditional models, AI models, and GenAI models with lifecycle tracking and automated documentation.

Product-wide claim; public pricing and bank-specific configuration remain unknown

Holistic AISource record

Holistic AI Identify module describes continuous discovery and centralized inventory for models, agents, APIs, datasets, and shadow AI across 15+ read-only connectors.

Product-wide claim; public pricing and SR 26-2 bank mapping are not published here

Lumenova AISource record

Lumenova AI describes registering AI agents, MCP servers, skills, and prompts with directory-based identity, holding unregistered agents detected in gateway traffic.

Covers gateway-routed agent traffic; traditional statistical model inventory and pricing are not published here

OneTrustSource record

OneTrust describes discovering and inventorying AI systems, models, agents, datasets, vendors, and use cases with lifecycle ownership and dependency mapping.

Suite product claim; plan tiering, module bundling, and pricing are not published here

CollibraSource record

Collibra docs (Nov 6, 2026) describe AI Command Center (evolved from AI Governance) centralizing AI Use Cases, Models, and Agents with 14 out-of-the-box platform integrations (Anthropic AI, AWS Bedrock AI, AWS SageMaker AI, Azure AI Foundry, Microsoft Copilot Studio, Microsoft Agent Builder, Azure ML, Databricks AI, Dataiku, Gemini Enterprise Agent Platform, MLflow AI, OpenAI, SAP AI Core, and Snowflake Cortex AI).

Requires Collibra permissions/edition; public pricing and bank SR 26-2 mapping unknown

Microsoft PurviewSource record

Microsoft Purview describes unified data security, DLP, Insider Risk, Compliance Manager, and Purview & Agent 365 visibility over AI agent data access, offered via Microsoft 365 E5, Purview Suite, and PAYG.

Data-security and compliance suite rather than standalone bank model-risk validation; unit rates require Azure/M365 pricing lookup

Fiddler AISource record

Fiddler AI describes maintaining a living inventory of every connected agent with user and team action, token, and cost tracking.

Agent observability/control inventory; traditional statistical model inventory and pricing remain unknown

DataRobotSource record

DataRobot describes tracking every asset and activity across the agent lifecycle for visibility over what is running where.

Agent workforce visibility claim; traditional bank statistical-model inventory fields and pricing remain unknown

Review workflows, audit trails, and exports

Includes VerifyWise (AG-MAN-004) and Credo AI (AG-MAN-007) workflow statements across 9 reviewed vendors. VerifyWise does not specify Trial inclusion for every workflow feature; Credo AI marks selected CI/CD, CASB, and API-gateway enforcement integrations as Planned.

VerifyWiseSource record

FAQ describes owners, contributors, reviewers and approvers; versioned attachments; and exportable audit reports. Pricing text does not securely bind each feature to Trial.

Do not infer Trial entitlements from extracted table text

Credo AISource record

Product page describes approval gates, evidence generation and audit trails; selected CI/CD, CASB and API-gateway integrations are marked Planned.

Not demonstrated behavior or implementation

ValidMindSource record

ValidMind describes configurable approval workflows, separation of duties across first, second, and third lines of defense, stress/backtesting validation, and audit-ready evidence.

Vendor product claim; plan gating and pricing are not published on this page

Holistic AISource record

Holistic AI Enforce module describes configurable sign-offs, remediation tracking, full audit trails, version history, and on-demand reports.

Vendor claim; deployment options and plan pricing remain unknown

Lumenova AISource record

Lumenova AI logs each decision with rule, reason, exact input, and version history exportable as CSV or JSON.

Vendor product claim; pricing and formal security certifications are not listed on this page

OneTrustSource record

OneTrust describes configurable intake, approval, attestation, and signoff workflows with lifecycle checkpoints and a continuous audit trail.

Vendor workflow claim; pricing and deployment terms remain unknown

CollibraSource record

Collibra lifecycle tracker assigns Owner and Assignees roles on assessments, automatically links models/agents to use cases, and tracks sign-off approvals.

Requires Collibra governance edition/permissions; export formats and pricing unknown

Fiddler AISource record

Fiddler AI describes inline request/response PII/PHI/secrets enforcement and audit evidence recorded for GRC standards and AI regulations.

Runtime observability and control focus; traditional bank MRM workflows and pricing unknown

DataRobotSource record

DataRobot describes enforceable access and approval controls, testing frameworks, and automated audit documentation across the agent lifecycle.

Broad agent workforce platform; plan entitlements, pricing, and bank MRM mapping unknown

← Return to requirement matrix

Make the next meeting useful

Build your demo brief.

Start with the questions your bank needs answered.

What belongs in an inventory?
Your meeting agenda
  1. Show a traditional model, a generative system and a third-party solution in the inventory.
  2. Run a submission, an independent challenge and an approval. Export the history.
Compare the workflow claims →