Use Case overview
A use case represents a specific application of an AI model (deployment) within a workspace. Define risk classifications aligned with the EU AI Act and track your use case through lifecycle stages. Filter use cases based on their lifecycle stage or risk classification from the use cases page.
Manage use cases
Add new use cases and assign Deployments to them in two ways:
- Manual: Create a use case on the use cases page first, then assign it when you create or update your Deployment
- Automatic: Create a Deployment without selecting an existing use case and Deeploy automatically creates one using your Deployment's name
All use case names must be unique within your Workspace, and you can't delete a use case that contains Deployments.
Use case details
Lifecycle stage progression
Track and advance your use case through four stages:
- Exploration
- Development
- Validation
- Production
Progress through stages in a disciplined manner. Move one step forward or backward at a time, without skipping stages:
- Permitted: development → validation, production → validation
- Not permitted: exploration → validation, production → exploration
Review status
Once the review period defined for the Workspace has passed, a use case's status automatically updates to 'Needs review'. This makes it easy to identify which use cases require attention and keep your governance process on track. After completing your review, open the use case and select Confirm review to update the status back to 'Reviewed'.
Risk classification
Assign a risk classification to a use case. The available classifications are based on the EU AI Act. Use the risk classification assessment to help determine the appropriate classification for your use case.
Role
Optionally assign a role to a use case based on your organization's responsibilities as defined in the EU AI Act. Choose Provider, Deployer or both.
The role assignment only affects the controls displayed for the use case when the EU AI Act control framework is applied to your Workspace. Different roles have different obligations under the EU AI Act, which are reflected in the applicable controls.
View Deployments
View all active Deployments part of the use case in one centralized location.
Controls
If a control framework is applied to the Workspace, Deeploy displays a list of controls. The list includes controls relevant to the use case risk classification, role and lifecycle stage. For example, if the use case is in the validation stage, controls relevant to both the exploration and validation stages appear.
Select View all controls to see all controls for this use case, including those for future stages. You can complete or provide evidence for future stage controls in advance, but they won't count towards your framework completion.
Evidence
Provide evidence that your use case meets the control. You can use several types of evidence:
- Checks automatically confirm when actions that contribute to a control’s implementation are complete
- Links let you reference evidence stored externally
- Images allow you to upload visuals, such as diagrams or screenshots
- Text lets you add free-form written evidence
Checks are defined at the control level. You can add other evidence types using the respective buttons. To edit or delete them, click the Options button.
Completing controls
To change the control status:
- Submit supporting evidence to automatically change status to 'In progress'
- Select Mark as completed to complete a control
- Select Mark as in progress to reopen a completed control when needed
Only completed controls count towards the use case framework completion.
Controls can be completed irrespective of the use case stage they belong to
Documentation
Add as many instances of documentation based on documentation templates or upload PDFs that are relevant for your use case. Learn more about Documentation Templates
PDF files up to 25 MB are accepted
Generate answer suggestions
This feature is currently exclusively available for the AI vendor assessment template. Use AI to generate suggested answers to template questions based on imported documents. We'll analyze imported documents to find relevant information and suggest appropriate answers for completing the template, saving time and ensuring comprehensive coverage. To use this feature:
- Add the AI vendor assessment template to your use case
- Click Generate suggestions.
- Select relevant imported documents
- Suggestions will load per question
- Open a question to view the suggestion
- Click Insert suggestion to populate the answer with the suggestion. You can also make additional edits. Alternatively, click Discard to ignore the suggestion.
The suggestions are generated by retrieving relevant text from documents for different questions using an embedding model and finally generating an answer using a LLM.
For saas users we use openAI 4o-mini and text-embedding-3-small to analyze and create embeddings of your uploaded document. Private cloud users can configure their own models for enhanced control at model level. In both the cases data is not stored or used for anything such as model training etc.
For private cloud installations customers need to enable AI feature in their installation.