Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
[This article is prerelease documentation and is subject to change.]
Important
This is a preview feature.
This preview feature is supported in production environments, but we recommend that you experiment and test thoroughly in a sandbox environment first. These features might have restricted functionality and are subject to supplemental terms of use.
This article is primarily intended for product owners, domain experts, consultants, and developers who want to experiment with AI capabilities in Business Central using the AI development toolkit.
Agents in Business Central can run on different AI models. The model an agent uses can affect its behavior, performance, and results. Multiple models can be available at the same time, and each agent can either use a specific model or rely on the automatic (default) model selection.
Note
The AI models available for agents in Business Central are determined by Microsoft and may change over time. Model availability can depend on factors such as the agent's capabilities, the environment, and the stage of the model lifecycle.
Where to see and control models
You can view and manage AI models in two ways:
- Agent configuration in the UI
For agents created from the agent design experience, the agent card and setup pages show the current model and allow you to change it. A dedicated list page shows all available models. For coded agents, the model in use isn't visible in the UI. - Agents coded in AL
APIs allow you to list available models and set the model on an agent or on individual tasks programmatically.
Model resolution
When an agent runs a task, Business Central resolves which model to use in the following order:
- If a model is specified on the task, that model is used.
- If no model is set on the task but a model is specified on the agent, the agent-level model is used.
- If no model is set on either the task or the agent, the default model is used.
This resolution order lets you set a general model preference at the agent level while overriding it for specific tasks that benefit from a different model.
Model lifecycle
Models follow a lifecycle similar to code deprecation in Business Central:
Preview → Available → Obsolete → Retired
The following points apply:
- Multiple models can coexist in the same environment.
- The model lifecycle doesn't align with the standard Business Central deprecation timelines.
- The default model (used by auto mode) can change over time.
- Agents that use auto selection might experience behavior changes when the default model changes.
- Model updates and the availability of new model-selection features in the UI can roll out independently. During transitions, agents continue to work as expected with the updated model, even before the selection UI reflects the change.
Important
When the default model changes, we recommend that you re-evaluate your agents, because agent behavior and accuracy can be affected by the model update.
Retired and obsolete models
When a model becomes obsolete or retired, the following indicators help you detect and respond to the change:
- UI indicators
Annotations for retired or obsolete models are shown on the agent card and model list. - Developer communication
Model lifecycle changes are communicated through AL Home notifications, documentation updates, and telemetry signals.
Related information
Configure agents programmatically (preview)
Coding agents in AL (preview)
Define and register an agent programmatically (preview)
Overview (preview)