AI & LLM Transparency
1. Overview
Involv Intranet includes optional AI-powered features ("AI Features") that enhance search, content creation, and knowledge discovery within your organisation's Microsoft 365 environment. This page explains how these features work, which AI models are involved, and how your data is protected.
Key principle: Involv AI Features run entirely within your organisation's own Microsoft Azure environment. Cognit does not process, access, or store any data handled by the AI Features.
2. Architecture — Customer-Controlled AI
Unlike many SaaS products that route data through the vendor's AI infrastructure, Involv takes a fundamentally different approach:
| Aspect | How it works in Involv |
|---|---|
| AI infrastructure | Your organisation's own Microsoft Azure OpenAI service, deployed within your Microsoft 365 tenant. |
| Data processing | All AI queries and responses stay within your Azure environment. Data never leaves your tenant. |
| API keys & credentials | Managed by your IT team in Azure. Cognit has no access to your Azure OpenAI endpoints or keys. |
| Model selection | Your IT team deploys the model version in Azure OpenAI. Involv connects to whichever model your organisation provisions. |
| Cost | AI usage is billed by Microsoft directly to your organisation's Azure subscription. |
3. AI Features Available
3.1 Intelligent Search
Enhances SharePoint search with natural language understanding. Users can ask questions in plain language and receive AI-summarised answers drawn from your organisation's content. The search index and results stay within your Microsoft 365 tenant.
3.2 Content Recommendations
Suggests relevant pages, news, and documents based on user context and content similarity. Recommendations are generated using your organisation's content graph — no external data is used.
3.3 Text Generation & Summarisation
Helps users draft, summarise, or rewrite content within the intranet. All generation happens through your Azure OpenAI service. Generated content is presented as a suggestion — users must review and approve before publishing.
4. Data Protection Commitments
| Commitment | Detail |
|---|---|
| No training on your data | Cognit does not use any customer data, AI queries, or AI outputs to train, fine-tune, or improve AI models. Microsoft's Azure OpenAI service also does not use your data for model training (see Microsoft's data privacy commitment). |
| No data extraction | Cognit cannot see, access, log, or extract AI prompts, queries, or responses processed through your Azure OpenAI service. |
| No sub-processor involvement | Because AI processing happens in your Azure tenant, there are no additional sub-processors for AI Features beyond Microsoft (your existing cloud provider). |
| GDPR compliance | AI Features process data under your organisation's existing data processing arrangements with Microsoft. No separate DPA is required for AI Features. |
| EU AI Act alignment | Involv's AI Features are classified as minimal or limited risk under Regulation (EU) 2024/1689. We maintain transparency about AI involvement (this page) and ensure human oversight (users review all AI outputs). |
5. EU AI Act Classification
Under the EU AI Act (Regulation 2024/1689), AI systems are classified by risk level. Involv's AI Features fall into the limited risk category:
| Requirement | How Involv complies |
|---|---|
| Transparency | Users are informed when content is AI-generated (visual indicators in the interface). This transparency page is publicly available. |
| Human oversight | All AI-generated content is presented as a suggestion. Users must explicitly approve, edit, or reject before publishing. |
| Technical documentation | This page serves as the required transparency disclosure. Additional technical details are available upon request. |
6. Enabling or Disabling AI Features
AI Features are optional and must be explicitly enabled by your organisation's IT administrator. If your organisation does not provision an Azure OpenAI service or does not enable the AI Features in Involv's configuration, no AI processing will occur.
Your IT administrator can disable AI Features at any time without affecting any other Involv functionality.
7. Related Documents
- Privacy Statement — how Cognit handles personal data
- Data Processing Agreement — processing terms for support and implementation scenarios
- General Conditions — Article 2.5 covers AI Features in the contractual terms
- EULA — Article 4.3 covers end-user responsibilities for AI Features
8. Contact
For questions about AI in Involv, contact:
Cognit BV — Data Protection
Email: [email protected]
Phone: +32 2 669 05 80
For general enquiries: Contact page
