Skip to main content
AlexO_WÜ
New Member ⭐⭐⭐
September 16, 2026
Solved

How does TOPdesk ensure that the AI LLM can only use customer data from my own TOPdesk system?

  • September 16, 2026
  • 2 replies
  • 50 views

How does TOPdesk ensure that the AI LLM can only use customer data from my own TOPdesk system?

 

How does TOPdesk prevent the LLM from using data from other TOPdesk systems or sending questions and answers to a global LLM?

 

Is the new LLM, “Robin,” capable of learning, or can it be trained?

Best answer by Sanne van Opstal-Brakel

Hi ​@AlexO_WÜ 

I asked your questions to or product manager of AI and got the following answers for you:

 

How does TOPdesk ensure the AI/LLM only uses data from my own TOPdesk system?

  • Your data is processed within your own tenant, on Microsoft's cloud infrastructure. No mixing across tenants or platforms.
  • Only the minimum data needed for a specific request is sent (your prompt, relevant context, system parameters).
  • The LLM infrastructure itself is scoped as part of your tenant setup, within the overall multi-tenant architecture.

How does TOPdesk prevent the LLM from using data from other TOPdesk systems, or sending your Q&As to a global LLM?

  • AI Features run on Microsoft Azure via the Microsoft Foundry Catalog. TOPdesk doesn't operate a shared global LLM that pools data across customers.
  • TOPdesk does not permit your prompts, completions, embeddings, or data to be used to train, retrain, or improve any shared foundation model, without your explicit permission.
  • TOPdesk doesn't share your prompts or outputs with any external model provider outside Microsoft Foundry.
  • Data residency: EU customers are processed within Europe, US customers within the US.

Is Robin capable of learning, or can it be trained?

  • Robin isn't retraining itself. It retrieves information intelligently based on your own data at query time.
  • AI Triage works the same way: retrieval and classification against your own closed ticket history, not training itself on that data.
  • Any improvement over time comes from your closed-ticket data getting richer, not from the model changing.

I hope this helps, if you have more questions, just let us know!

2 replies

Ayman
Starter ⭐⭐
September 16, 2026

Hi Alex,

How does TOPdesk ensure that the AI LLM can only use customer data from my own TOPdesk system?

Not too much details but I read that the isloation is done through the an MCP Layer, you can review the documentaion and this link: https://tip.topdesk.com/c/200-ai-mcp-based-service- 

 

How does TOPdesk prevent the LLM from using data from other TOPdesk systems or sending questions and answers to a global LLM?
Is the new LLM, “Robin,” capable of learning, or can it be trained?

 

What’s interesting is: probably this is the first honest AI chat bot that might tell you “I don’t know”

“Robin uses exclusively these controlled sources and does not browse the open internet. If a reliable answer cannot be found, Robin will simply indicate that, without guessing….Then the chatbot offers to create a support ticket for you directly”
this actually made me smile.

    Sanne van Opstal-Brakel
    Employee
    Sanne van Opstal-BrakelAnswer
    September 22, 2026

    Hi ​@AlexO_WÜ 

    I asked your questions to or product manager of AI and got the following answers for you:

     

    How does TOPdesk ensure the AI/LLM only uses data from my own TOPdesk system?

    • Your data is processed within your own tenant, on Microsoft's cloud infrastructure. No mixing across tenants or platforms.
    • Only the minimum data needed for a specific request is sent (your prompt, relevant context, system parameters).
    • The LLM infrastructure itself is scoped as part of your tenant setup, within the overall multi-tenant architecture.

    How does TOPdesk prevent the LLM from using data from other TOPdesk systems, or sending your Q&As to a global LLM?

    • AI Features run on Microsoft Azure via the Microsoft Foundry Catalog. TOPdesk doesn't operate a shared global LLM that pools data across customers.
    • TOPdesk does not permit your prompts, completions, embeddings, or data to be used to train, retrain, or improve any shared foundation model, without your explicit permission.
    • TOPdesk doesn't share your prompts or outputs with any external model provider outside Microsoft Foundry.
    • Data residency: EU customers are processed within Europe, US customers within the US.

    Is Robin capable of learning, or can it be trained?

    • Robin isn't retraining itself. It retrieves information intelligently based on your own data at query time.
    • AI Triage works the same way: retrieval and classification against your own closed ticket history, not training itself on that data.
    • Any improvement over time comes from your closed-ticket data getting richer, not from the model changing.

    I hope this helps, if you have more questions, just let us know!

    Sanne van Opstal-Brakel, your TOPdesk Community Fairy Godmother
      Terms & ConditionsAccessibility statement