AI Engine connects to one AI provider at a time. You choose it in the Org Profile section of the Ai Agent Admin Setup tab. This page lists the supported providers, how AI Engine calls them, and which models you can pick.
Providers#
| AI Provider option in the Org Profile | API AI Engine uses | Named Credential URL | AI Model field |
|---|---|---|---|
| OpenAI (Responses API) | OpenAI Responses API | https://api.openai.com |
Chosen from a list |
| Anthropic Claude | Anthropic Messages API | https://api.anthropic.com |
Chosen from a list |
| Azure OpenAI | Azure OpenAI chat completions on your deployment | https://YOUR_RESOURCE.openai.azure.com |
The name of your Azure deployment, typed in |
When you switch AI Provider, the Org Profile fills in a default model and a default Named Credential name (OpenAI, Anthropic or AzureOpenAI). After switching:
- Check that the Named Credential field matches the name of the Named Credential you created in Setup. Change it if yours is different.
- On Azure OpenAI, replace the prefilled
gpt-4oin AI Model with the name of your own deployment. - Save the profile and ask a test question in the chat.
Creating the Named Credential itself, including adding CloudConnected to its allowed namespaces for callouts, is described in Connect your AI provider.
Models#
For OpenAI and Anthropic, AI Model is a list of the models the current AI Engine version supports, each with a short note such as "recommended", "top quality", "fast & cheap" or "experimental". The recommended defaults are:
| Provider | Default main model | Default fast model |
|---|---|---|
| OpenAI | GPT-4.1 (gpt-4.1) |
GPT-4.1 mini (gpt-4.1-mini) |
| Anthropic | Claude Sonnet 4.6 (claude-sonnet-4-6) |
Claude Haiku 4.5 (claude-haiku-4-5) |
| Azure OpenAI | Your deployment name | Same as the main model, unless you enter a second deployment |
The list is updated with new package versions. Models marked experimental are available to try; test them on your own data before switching production users. For guidance on picking, see Choose the main and fast models.
Main model and fast model#
The Org Profile has two model settings:
- AI Model (the main model) handles chat, record actions, Meet Prep, AI scoring, PDFs and Magic Snap.
- Fast model (internal steps) handles the web-search steps (Web Check, Company News, web research in the chat, Meet Prep web search, Verify Company On The Web, Magic Snap research and Find companies) and Help mode, including the assistant on the admin Help & Setup tab. Turning a chat question into a query uses the main model.
The fast model must come from the same provider as the Named Credential; a model from another provider is not saved. If you leave it on Provider default, web-search steps use the default fast model from the table above and Help mode uses the main model.
On Azure OpenAI, leave the fast model blank to use the main deployment, or enter the name of a second, cheaper deployment. Because web search on Azure goes through OpenAI (see below), the Azure fast model affects Help mode only.
Provider-specific behaviour#
- Temperature is sent only to models that accept it. AI Engine leaves it out for models that reject it, such as the GPT-5 family and o-series on OpenAI and the newer Claude models.
- Web search uses the provider's own web search tool on OpenAI and Anthropic. Azure OpenAI does not offer it, so an Azure-configured org sends web searches to OpenAI through a Named Credential named exactly
OpenAI, usinggpt-4.1-mini. To use web features on Azure, create that Named Credential as well. If your data must stay within Azure, keep in mind that these features send data to OpenAI. See Choosing a provider for compliance. - Images (Magic Snap) are supported on all three providers.
- Azure deployment names may contain only letters, numbers, dots, hyphens and underscores.
After changing a model#
Models differ in how reliably they return structured results. After any model change, test Meet Prep, AI Score and answers shown as cards before users rely on them.