Running/training Document Intelligence Custom Layout model on-prem

Hi Elle MacLennan,

Good question about Azure AI Document Intelligence and deploying a custom layout model on prem thanks for sharing your scenario. From what you described:

• You want to train a custom layout model in the cloud (with redacted documents)
• Then you’d like to deploy that model on-premises

  What the service currently supports

• The training of a custom layout model for Document Intelligence happens in the Azure cloud. That is, you use the Azure service (which is cloud-hosted) to build the model.
• The documentation does not mention a supported path for taking that model and running it fully on-premises (i.e., in your own data centre, disconnected from Azure).
• So in simple terms: training in cloud = supported; on-premises deployment of the trained model = not currently a standard/advertised option.

  What it means for your scenario

• If you train your redacted-document layout model in the cloud, you’ll be operating within the supported flow.
• But then wanting to deploy the same custom model on-premises likely means you’ll hit limitations: the runtime, infrastructure or licensing may require the Azure cloud service.
• If “on-premises” means you need the model to run entirely inside your own network (without relying on Azure cloud endpoints), you should assume that may not be supported (at least not yet) or you might need a special arrangement/licence or hybrid setup.

  What you can do next

• Check the official Microsoft docs or reach out to your Microsoft rep to confirm: Is a “bring your own model / run on-prem” feature available for custom layout models?
• If cloud-only: consider whether you can keep your processing in Azure (maybe with strong network/security controls) instead of fully on-premises.
• If on-premises is a must (due to data sovereignty, offline requirement, etc), look for alternative solutions or hybrid modes (maybe an Azure stack variant, or edge/cloud hybrid).
• Review the SLA, connectivity and authentication when your model stays in cloud but your data reside on-premises – ensure this meets your security/compliance needs.

In short: you can train a custom layout model in Azure’s cloud for Document Intelligence. Deploying that same model wholly on-premises is not clearly supported (at least based on the public documentation). If you really need on-prem deployment, you’ll want to verify directly with Microsoft or explore hybrid/edge options.

References Documents:

• What is Azure AI Document Intelligence?
• Document Intelligence Custom Models
• What is the Document Intelligence Layout Model?

Hope this helps clarify things! If you feel that your quires have been resolved, please accept the answer by clicking the "Upvote" and "Accept Answer" on the post.

Thank you!