Activation of imc AI Services

Disclaimer text 

AI-enhanced content search - compliance information 

  

General Information about the search 

With the product version 14.25.01 imc  integrates a new platform-search functionality (Elasticsearch) into the imc Learning Suite.  Elasticsearch offers two distinct search modes (or types), which differ technically: 

• Lexical Search
  This mode directly matches the words from the user's query with titles, descriptions, and keywords of the indexed learning content. It’s a purely text-based lookup — no artificial intelligence involved. 
• Semantic Search
  In this mode, both the indexed learning content and the user’s query are converted into vectors using AI technologies. This allows the system to detect semantic similarities and retrieve related concepts and themes even if they are phrased differently from the query. 

Customers can choose between these two modes per system. Lexical search is included out-of-the-box with the imc Learning Suite and is available by default. To use semantic search, the required AI services must be activated by imc via Service Desk. 

 For more details on the search, its functions, how to configure and use it, please click here.

Third-Party Providers 

The following AI services are integrated into the imc Learning Suite to enable the search: 

• Elasticsearch: standalone search and data analysis service using searchable indices accessed via REST APIs. 

• Ollama: open-source tool for locally running large language models (LLMs) to support text generation and embeddings. 

• Jeffh: embedding model used. 

• Grafana: monitoring platform used by imc for analytics. 

 

Access Control (ACL) 

Search results fully respect the access permissions defined by the Customer. Each user sees only the learning media, courses, and learning paths they are authorized to access. Therefore, results may vary between users. 

  

Hosting  

All services for the search are running on the imc Clouds. 

imc does not develop proprietary AI models. Instead, it provides third-party AI services integrated into the imc Learning Suite. All data processing related to these services takes place solely within imc’s Azure Cloud infrastructure.  

 

Activation/Deactivation 

The search will be delivered within the standard product to all Customers with default setting “lexical search" (no AI involved). The search becomes visible to users once Customers] have configured the navigation points (Configuration details).

To activate the Semantic search, Customers need to have read and agreed to terms and description of the AI services included on this page. Afterwards imc Service Desk will coordinate its activation on the desired system.  

Please note, after switching to the Semantic search, a full re-indexing should be carried out, as the embedding service must now vectorize the entire content available on the platform.  

At any time, Customers may decide to discontinue the use of AI services and revert back to lexical search: 

• Using the switch in the search configuration admins can change the search type back to lexical. This will change the method by which the search results are retrieved and ranked for the users (as explained at the beginning of this document).  

• Please note that switching back to lexical search will not automatically de-activate nor revert the deployment of the AI services. If Customers wish to have the AI services removed, this can be requested via imc Service Desk. 

Fair Use Policy 

The semantic search service is generally available to Customers without fixed usage limits. However, the provider expects fair and appropriate use within the scope of the contractually agreed purpose. 

Usage that significantly exceeds the average of comparable Customers, or that negatively impacts system performance, stability, or availability for other users, may be classified as excessive. In such cases, the provider reserves the right to contact the Customer to jointly agree on appropriate measures—such as technical adjustments, usage restrictions, or a transition to a higher-performance model. 

Any restrictions will only be implemented following prior consultation. 

The provider also reserves the right to update this Fair Use Policy as needed. Customers will be informed of any changes in a timely manner. 

  

Data Protection 

To provide and improve the search functionality, imc processes certain metadata, technical data, and usage-related information. While these data categories are not inherently personal, it cannot be fully excluded that such data may, in certain cases, contain or permit a link to personal data. Accordingly, imc treats such data in accordance with applicable data protection laws. 

 

Legal Basis 

The processing of such personal data in this context is carried out: 

pursuant to Art. 6(1)(b) GDPR (performance of a contract), insofar as the processing is necessary for providing the agreed platform functionalities, including the search capabilities selected by the Customer; 

pursuant to Art. 6(1)(f) GDPR (legitimate interest), namely to provide relevant search results and improve system performance and user experience.  

imc ensures that such processing does not override the fundamental rights and freedoms of the data subject. 

 

Data Minimisation and Retention 

Search queries and related technical data are processed only to the extent necessary to provide the service, ensure system security, and optimise performance. Data is retained only as long as required for these purposes and is deleted in accordance with applicable data protection provisions. 

 

No Use for Model Training 

Customer data and user input are not used to train, fine-tune, or otherwise develop machine learning models. 

 

Processed Data Includes: 

Subscription Service 

• Embedding subscription and usage volume per tenant, group and user (anonymously) are saved in an imc-managed database.  

Indexing & Embeddings 

• Metadata from indexed Customer content (titles, descriptions, keywords) are stored in the search index. 

• Embeddings for both search queries and content are created using the embedding service within the Azure environment; data does not leave the imc-controlled infrastructure. 

User Search Queries 

• Search inputs and recent search history (e.g., last six queries) are stored in an Azure Cloud database. 

• Users see their own search history in the UI. They can delete it at any time.  

• If semantic search is activated: imc tracks the amount of used word embedding tokens used per user (anonymously), per group and per client for search queries. 

• Metrics and KPI
  Stored metrics are not user-related and include: 

• Timestamp of the search. 

• Query content. 

• Number of queries per Customer and tenant. 

• Number of results returned . 

• Number of zeroresult searches. 

• Relevance score of results. 

• Possibly in the future: active filters used in queries.  

 

• Search request profiling: user inputs may be stored as long as necessary to improve system functionality, troubleshoot errors, or analyze usage patterns; data is deleted according to legal requirements once no longer needed. 

 

General Notes 

• Neither search inputs nor indexed Customer data are used for training or optimizing machine learning models; no further processing is done for model development. 

• Customers are strongly advised to inform their users not to enter sensitive, confidential, or personal data in the search field. 

• Use of the search function is at the user’s own risk. imc AG assumes no liability for any harm resulting from data input, transmission, processing, or misuse by third parties. 

• By using the search function, users accept responsibility for the content they provide and waive any claims against imc AG regarding its use or processing. 

 For usage guidelines and recommendations, consult this page.