New: Bring Your Own Model – Azure OpenAI

With this release, EnateAI Integrations can now run their inference on your own Azure endpoints instead of Enate's. The prompt chain, agent design, Integration behavior and Enate's quality underwriting are all identical, the only thing that changes is where the model call executes. You get the same EnateAI quality you already rely on, while the model call itself and your data stays inside your own Azure subscription.

What is Bring Your Own Model?

Instead of Enate routing a Integration's model calls to Enate's own Azure-hosted OpenAI instance, you can supply your own Azure-hosted endpoint for that same Integration to call instead. Everything else carries on exactly as it did before.

Security Note: The key provided to configure your model is securely stored within Enate's Azure Key Vault. All credentials are encrypted at rest and protected using industry-standard security controls to ensure only authorized services can access them

Setting up BYOM for a EnateAI Integration

First navigate to Builder and go to Marketplace and select the EnateAI Integration you want to configure. Then click Activate. A pop-up will appear giving you the choice to either Activate with Enate AI or to connect to your own endpoint instead.

If you want to use your own endpoint, fill in the three connection details requested (your Azure endpoint URL, key, and the deployment name of the model you've provisioned).

For EnateAI IDP Integrations, you will need to fill out the below pop-up.

Once you have populated these details, click Test Connection. This checks two things: that Enate can reach your endpoint, and that it's serving the correct Azure model version for that Integration. Once the test passes, click the Activate button and the EnateAI Integration will activate connecting to your endpoint. It should be noted that you can not select to activate until you have a successful test.

If you're sticking with Enate-managed model, just leave the pop-up blank and hit Activate with EnateAI as normal.

How do you update Integration details for BYOM?

To update the details i.e API Key, go to Marketplace and find the Integration you need to update. You will see that there is an Update button. Clicking this button will bring up a pop-up, where you can input updated details and then click Test Connection.

Once you have a successful test, simply click Update and Activate.

What if you already are using an Integration with EnateAI?

If you are currently using a EnateAI Integration with EnateAI's own endpoint and wish to change to using your own endpoint, all you need to do is click on the update button on the Integration. This will then bring up a pop-up stating that "This adapter is currently active with Enate's managed model". Simply fill out the required details, test the connection and then update and activate.

EnateAI Document Classification and IDP

EnateAI Document Classification and IDP need more than one underlying AI model (an Azure OpenAI endpoint and an Azure Document Intelligence endpoint). For these, BYOM only takes effect once every required endpoint is verified. Registering just one of the two keeps the whole Integration running on Enate-hosted inference.

Endpoint requirements

Your endpoint has to be a genuine Azure host (Azure OpenAI, Azure Document Intelligence) serving the model family that Integration requires this is what keeps Enate's Quality Underwriting intact. Native OpenAI or other non-Azure endpoints are rejected at registration and will not work.

The exact Azure model version required changes over time as we upgrade the AI behind a Integration. Please check the EnateAI page to find what models are currently in use.

Keeping your configuration current

Model version changes: Before Enate upgrades a EnateAI Integrations model, Enate will update the API configuration to allow you to prepare for the next model while still working with the existing model. If Enate updates the model and you have not provided the new endpoint configuration for a specific Integration, the Integration will fall back to Enate-hosted Model until you update your endpoint details for the new model.

BYOM is rolling out across our production EnateAI Integrations with this release. Integrations still in Beta (such as AI Analyst) will get BYOM support in a later release.