AI Governance operating model

The AI Governance operating model includes the following AI Governance-specific asset types:

Asset type Description UUID
AI Model An AI Model is a statistical model that is trained on datasets in order to make predictions on new data. 00000000-0000-0000-0000-000000031402
AI Use Case An AI Use Case is a specific application of artificial intelligence (AI) and AI Models trained on specific data in order to solve a business problem and deliver business value. Implementing an AI Use Case may result in automating a task, improve decision-making or develop new products and services. 00000000-0000-0000-0000-000000031401

The following image shows the relations between the AI Governance asset types and other relevant asset types.

For instructions on how to create this diagram view in your Collibra environment, see Create an AI Governance operating model diagram view, at the bottom of this topic.

Descriptions, attributes, and relation types

Click the relevant tab below to see descriptions, attributes, and relation types for each asset type.

  • AI Model
  • AI Use Case

Description

A statistical model that is trained on datasets in order to make predictions on new data.

Relation types

Relation type Head role / corole tail Public ID
used in AI Use Case AI Use Case uses / used in AI Model AIUseCaseUsesAIModel
trained by Asset AI Model trained by / trains Asset AIModelTrainedByAsset
infers from Asset AI Model infers from / used to infer Asset AIModelInfersFromAsset
has output Asset AI Model has output / is output Asset AIModelHasOutputAsset
is provided by Vendor AI Model is provided by / provides Vendor AIModelProvidedByVendor
uses AI Model AI Model uses / is used by AI Model AIModelUsesAIModel
complies to Governance Asset Asset complies to / applies to Governance Asset AIAgentUsesAIModel

Attributes

Attribute Description Public ID
Description The description of the asset. This is typically a more verbose way to describe what the asset means. Description
Model Accuracy

Model Accuracy refers to how well the model performs on a given task. Typically, this is defined by the proportion of correct predications made by the model. For example, if a model is used to classify emails as spam or not spam, and it classifies 90% of emails correctly, then the model accuracy is 90%.

 ModelAccuracy
Model Precision

Model Precision refers to how accurate positive model predictions are. Typically, this is defined as the proportion of predictions that are correct. For example, if a model is used to classify emails as spam, and it correctly classifies 95% of emails as spam, then the model has a precision of 0.95.

 ModelPrecision
Mean Squared Error

Mean Squared Error (MSE) refers to a model quality metric that measures the quality of the model’s predications.

 MeanSquaredError
Mean Absolute Error

Mean Absolute Error (MAE) refers to a model quality metric that evaluates the performance of regression models.

 MeanAbsoluteError
Model Type

Type of AI model. Values include: Generative AI, Classification, Regression, Computer Vision, Reinforcement Learning, and Image Classification.

 ModelType
Retrain Cycle

The frequency with which the model is retrained.

 RetrainCycle
Feature Importance

Feature Importance refers to how important a feature is to a machine learning model. It helps you to understand which features contribute the most to the model’s predictions.

 FeatureImportance
Version If this asset is versioned (manually or in an external system), this string represents the asset version. Version
Repository Reference to the repository where the code behind the model is stored. Repository

Description

An AI use case is a specific application of AI and AI Models, trained on specific data in order to solve a business problem and deliver business value. Implementing an AI use case may result in automating a task, improving decision-making, or developing new products and services.

Relation types

Relation type Head role / corole tail Public ID
is assessed by Assessment Review Asset is assessed by / assesses Assessment Review

AssetIsAssessedByAssessmentReview
uses AI Model AI Use Case uses / used in AI Model

AIUseCaseUsesAIModel
infers from Asset AI Use Case infers from / used to infer Asset

AIUseCaseTransformsAsset
trained by Asset AI Use Case trained by / trains in Asset

AIUseCaseTrainedByAsset
has output Asset AI Use Case has output / is output Asset

AIUseCaseHasOutputAsset
complies to Governance Asset Asset complies to / applies to Governance Asset AssetCompliesToGovernanceAsset

Attributes

Attribute Description Public ID
Description

General description of the Use Case and its potential for the use of AI.

Description
Use Case Application

Indicates that the AI use case will be used by an external audience or internally by your organization.

UseCaseApplication
Use Case Stage The stage that your AI use case is in. For example: Ideation, Development of Data and Models, or Implementation and Monitoring. UseCaseStage
Business Case

Refers to the business problem you want to solve with the AI use case. It focuses on describing a concrete business problem. For example, I'm a customer support manager and my team receives too many support tickets.

BusinessCase
Business Value

Refers to how this AI use case can improve your organization. For example, reduce support tickets, bring in additional revenue, or mitigate risks.

BusinessValue
Business Sponsor

Refers to the Business Owner or Executive Sponsor of the AI Use Case in your organization.

 BusinessSponsor
Maintenance Cost

Indicates the overall expected cost of running the Use Case over selected period of time.

 MaintenanceCost
General Purpose AI

Refers to whether the model used in your AI use case is using a General Purpose AI (GPAI). Some regulatory frameworks may impose additional transparency and risk mitigation requirements for GPAI based systems, sometimes referred to as foundation models. For example, large language models (LLMs).

 GeneralPurposeAI
Third Party Model

Refers to the vendor of your AI model(s) and what kind of model you are using. For example, Google Vertex.

 ThirdPartyModel
Internal Model

Refers to the existing or upcoming internally built model(s) your AI use case may use.

 InternalModel
Training Data Description

The training or re-retraining data used to teach the AI model(s).

 TrainingDataDescription
Inference Data Description

An explanation of the input or inference data the AI model(s) uses to create output data. For example, an image classification model uses images as input data and a language model uses text as input data.

 InferenceDataDescription
Model Output

An explanation of the output data that the AI model(s) is expected to create. For example, classification labels, descriptions, or complex probability predictions.

 ModelOutput
Data Storage

Refers to whether any data is stored, and if so, where and how the data is stored. For example, the prompts data is stored on the cloud.

 DataStorage
Automation Level

The nature and degree of automation of the AI use case. For example, are decisions going to be based solely on the automated output, or is human oversight possible or planned?

 AutomationLevel
Model Monitoring

Refers to how your organization will ensure the AI model is meeting accuracy and performance expectations.

 ModelMonitoring
Legal Approval Date

Date your legal team approved or rejected the AI use case.

 LegalApprovalDate
Legal Approval Renewal Date

Date of the expected periodical review of the Use Case’s approval.

 LegalApprovalRenewalDate
Legal Description of Model

Description of the AI model provided by your legal team. This may include any legal repercussions of processing the AI model within the context of the AI use case.

 LegalDescriptionofModel
Security Protocols

Any general security protocols that may result from the implementation of the AI use case.

 SecurityProtocols
Data Retention Protocols

Any data retention standards already in place or projected to be implemented for the AI use case.

 DataRetentionProtocols
Data Privacy Risks

Any data privacy risks that may result from processing the AI model within the context of this AI use case and either putting it on the market for external customer use or internal use. For example, is there the risk of a data breach or misuse of data?

 DataPrivacyRisks
Data Privacy Risk Score Number calculated from Risk Assessment. DataPrivacyRiskScore
Intellectual Property Risks

Inherent Intellectual Property Risks resulting from processing AI Models within this use case and placing them on the market or putting into service for own use. Examples include copyright or patent infringement.

 IntellectualPropertyRisks
Ethical Risks

Any ethical risks that may result from processing the AI model within the context of this AI use case and either putting it on the market for external customer use or internal use. For example, is there the risk of a data breach or misuse of data?

 EthicalRisks
Other Risks

Other Risks resulting from processing AI Models within this use case and placing them on the market or putting into service for own use. Examples can include safety and reliability, security, social risks.

 OtherRisks
Business Risks A summary of the business risks associated with implementing the AI use case. For example, the expected financial loss resulting from significant disruptions to the business due to the complexity of operating the AI model. BusinessRisks
Overall Risk Analysis

Details and the result of any risk analysis performed on the AI use case.

 OverallRiskAnalysis
Overall Risk Rating

Risk level calculated based on pre-defined thresholds in the default Risk Assessment.

 OverallRiskRating
Transparency Disclosure Requirements

Transparency Disclosure Requirements refers to identified requirements, if any, for AI use case transparency. For example, a Transparency Declaration that requires an organization to disclose the purpose of the AI model and any data used for training purposes.

 TransparencyDisclosureRequirements
Protective Measures

Any additional required or recommended actions that have been identified based on regulations, industry standards or dedicated frameworks.

 ProtectiveMeasures

Asset statuses

Tip You can configure which initial status you want AI Use Case assets to have when they are registered, and the asset status progression – in other words, the lifecycle stages through which you want your AI use cases to evolve. For more information, go to Configure AI Use Case asset statuses and The lifecycle of an AI use case.

Status Description
Ideation The AI use case is just a concept, to be reviewed with regard to value and risks. It is not yet used in production. It can be considered a proof of concept.
Development

The AI use case is in, or is ready to enter, the development process.
Monitoring The AI use case in production, is actively used and monitored with regard to business value and risks. AI models are monitored with regard to performance, reliability, and accuracy.
Archived

The AI use case is no longer in production. It is archived for documentation and audit purposes.
Rejected After being assessed for its value and inherent risks, the use case was not approved and not put into production.

Tip For information on editing the status of an asset, go to Edit an asset.

Create an AI Governance operating model diagram view

You can create an AI Governance diagram view, to visualize the operating model. The following procedure describes how to quickly create a new diagram view by copying and pasting the JSON code in the diagram view text editor.

Steps

  1. Open an asset page.
  2. Click the Diagram tab.
    The diagram is shown in the default diagram view.
  3. Click to add a new view.
  4. Click the Text tab, to switch to the diagram view text editor.
  5. Click Show me the JSON code below this procedure, to expand the code.
  6. Copy and paste the code in the diagram view text editor.
  7. Click Save.
  8. Edit the name and description of the diagram view, to suit your needs.
{
    "nodes": [
        {
            "id": "AI Use Case",
            "type": {
                "id": "00000000-0000-0000-0000-000000031401"
            },
            "fields": [],
            "editorSettings": {
                "edgePropsExpanded": true,
                "nodePropsExpanded": false
            }
        },
        {
            "id": "AI Model",
            "type": {
                "id": "00000000-0000-0000-0000-000000031402"
            },
            "editorSettings": {
                "nodePropsExpanded": false,
                "filterPropsExpanded": true
            }
        },
        {
            "id": "Vendor",
            "type": {
                "id": "00000000-0000-0000-0000-000000031403"
            }
        },
        {
            "id": "Assessment Review",
            "type": {
                "id": "00000000-0000-0000-0000-000000031305"
            }
        },
        {
            "id": "Policy",
            "type": {
                "id": "00000000-0000-0000-0000-000000031202"
            }
        },
        {
            "id": "Asset",
            "type": {
                "id": "00000000-0000-0000-0000-000000031000"
            }
        },
        {
            "id": "Asset 2",
            "type": {
                "id": "00000000-0000-0000-0000-000000031000"
            }
        },
        {
            "id": "Asset 3",
            "type": {
                "id": "00000000-0000-0000-0000-000000031000"
            }
        }
    ],
    "edges": [
        {
            "from": "AI Model",
            "to": "Vendor",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007105"
            },
            "roleDirection": true
        },
        {
            "from": "AI Use Case",
            "to": "AI Model",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007098"
            },
            "roleDirection": true
        },
        {
            "from": "AI Use Case",
            "to": "Assessment Review",
            "label": "",
            "type": {
                "id": "00000000-0000-0000-0000-000000007097"
            },
            "roleDirection": true
        },
        {
            "from": "AI Use Case",
            "to": "Policy",
            "label": "",
            "type": {
                "id": "00000000-0000-0000-0000-000000007018"
            },
            "roleDirection": true
        },
        {
            "from": "AI Use Case",
            "to": "Asset",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007100"
            },
            "roleDirection": true
        },
        {
            "from": "AI Use Case",
            "to": "Asset 2",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007099"
            },
            "roleDirection": true
        },
        {
            "from": "AI Use Case",
            "to": "Asset 3",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007101"
            },
            "roleDirection": true
        },
        {
            "from": "Policy",
            "to": "AI Model",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007018"
            },
            "roleDirection": false
        },
        {
            "from": "AI Model",
            "to": "Asset 2",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007103"
            },
            "roleDirection": true
        },
        {
            "from": "AI Model",
            "to": "Asset",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007102"
            },
            "roleDirection": true
        },
        {
            "from": "AI Model",
            "to": "Asset 3",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007104"
            },
            "roleDirection": true
        },
        {
            "from": "AI Model",
            "to": "AI Model",
            "label": "",
            "style": "arrow",
            "type": {
                "id": "00000000-0000-0000-0000-000000007106"
            },
            "roleDirection": false
        }
    ],
    "showOverview": false,
    "enableFilters": false,
    "showLabels": true,
    "showFields": true,
    "showLegend": true,
    "showPreview": false,
    "visitStrategy": "directed",
    "layout": "HierarchyTopBottom",
    "maxNodeLabelLength": 50,
    "maxEdgeLabelLength": 30,
    "layoutOptions": {
        "compactGroups": false,
        "componentArrangementPolicy": "topmost",
        "edgeBends": true,
        "edgeBundling": true,
        "edgeToEdgeDistance": 5,
        "minimumLayerDistance": "auto",
        "nodeToEdgeDistance": 5,
        "orthogonalRouting": true,
        "preciseNodeHeightCalculation": true,
        "recursiveGroupLayering": true,
        "separateLayers": true,
        "webWorkers": true,
        "nodePlacer": {
            "barycenterMode": true,
            "breakLongSegments": true,
            "groupCompactionStrategy": "none",
            "nodeCompaction": false,
            "straightenEdges": true
        }
    }
}