Gen AI: AI Connector and Model Selection

Uncountable users with Gen AI configured can use AI connectors to control which AI backends power Gen AI features across the platform. Admins can add, test, and prioritize connectors, and individual users can switch models on demand from within supported AI features.

Configuring connectors requires admin access. Switching models on demand is available to all users when multiple eligible models are available.

How Model Selection Works

When you use any Gen AI feature, Uncountable automatically selects the right model for the task based on your configured priority order:

1. The system checks what capabilities are required (e.g. tool use, web search).
2. It selects the highest-priority model from your list that supports those capabilities.
3. If no available model supports the required capabilities, an error is displayed.

This means you can configure multiple models and let the system route automatically, or set a single preferred model for all features.

Configuring Available Connectors (Admin)

To access connector settings:

1. Open Company-Wide Settings.
2. Go to Global Permissions > Modules.
3. Select Configure Generative AI Module.
4. Open Manage Connectors.

From this page you can:

• Add connectors using guided setup for Uncountable-managed connectors, your own API key, or a custom model endpoint.
• Test connectors from the list before relying on them in production.
• Reorder connectors to change priority. The connector at the top of the list has the highest priority and is used first when it supports the required capability.
• Archive or disable a connector to prevent it from being used.
• Restrict available model families for AWS Bedrock connectors, so the account can limit AI use to specific vendors, regions, or model families.

If you use a Bring Your Own Model (BYOM) connector — connecting Uncountable to your own OpenAI, Anthropic, Google Gemini, AWS Bedrock, or custom model endpoint — those connectors appear in Manage Connectors and can be prioritized alongside Uncountable-managed connectors. Refer to Setting Up a Bring Your Own Model (BYOM) AI Connector for setup instructions.

Switching Models On Demand

When using Bodie AI, you can switch the active model directly from the sidepanel without leaving your current context:

1. In the Bodie AI sidepanel, click the … icon.
2. Select Select Model Preference.
3. Choose from the available models.

Your selected preference applies to your current Bodie AI session. The system will use the model you selected as long as it supports the capabilities required for the task.

Viewing AI Activity and Debugging

Each AI interaction includes a full trace you can inspect to understand what the model did and why.

To view activity for a specific interaction:

1. While using any AI feature, click the … icon.
2. Select View All Activity to open a sidepanel with a detailed trace showing:
   • User input
   • Tools the AI called
   • Results and outputs

To view all sessions:

1. From the activity sidepanel, click Options > View All Sessions.
2. This opens the Sessions listing page, which shows a complete log of AI sessions across the platform that you have permission to view.

This is useful for troubleshooting unexpected AI behavior, auditing AI usage, and understanding how specific responses were generated. If a colleague shares a Bodie chat with you, you can also view the underlying AI session for that shared conversation.