OpenAI
OpenAI Connections support selected language, image, file, and agent workflows. A deployment can use an iDialogue-managed provider account or an approved customer-provided API key.
This connection is enabled by iDialogue support. Contact support@idialogue.app to review requirements and enable an OpenAI connection for your org.
OpenAI connection support is available through the Governance tier.
What this connection does and does not do
The OpenAI Connection sends selected instructions and business context to the OpenAI API when an approved iDialogue workflow requires model processing.
It does not:
- expose a connected Salesforce org to public AI models;
- create a shared knowledge base across iDialogue customers;
- give the model standing Salesforce credentials; or
- use data from the standard iDialogue-managed integration to train OpenAI models.
Approved dialogue context can support continuity across turns. Document generation, OCR, image analysis, file Q&A, and transactional work remain task-based unless the workflow explicitly uses dialogue continuity.
Requested files and generated artifacts can be retained in the secure iDialogue repository or shared through an iDialogue Experience or Document Room. Customer responses, review decisions, and eSignature outcomes may update Salesforce through the configured workflow.
Choosing an OpenAI connection
iDialogue-managed connection
iDialogue manages the provider account, service administration, and billing needed to deliver the approved OpenAI-powered experiences. Customers continue to govern which agents, Skills, Salesforce data, files, and workflows are available in iDialogue.
Customer-provided key
A customer-provided key associates provider calls with the customer's configured OpenAI organization and project. It can give the customer provider-level administration, usage visibility, billing, rate limits, and access to provider data-control settings.
The key changes provider-level administration. It does not replace the iDialogue workflow controls described below.
Data controls and retention
Evaluate the OpenAI Connection as part of the complete iDialogue workflow:
- Model training: Customer data sent through the standard iDialogue-managed OpenAI integration is not used to train OpenAI models.
- Selected workflow context: The model receives only the instructions, files, business context, and tools supplied for the configured request.
- Dialogue continuity: Approved context can carry forward when the workflow is designed to continue a business conversation.
- Task processing: Document generation, OCR, image analysis, file Q&A, and transactional work do not automatically create dialogue history.
- Provider application state: Some provider features retain response state under the provider's API controls to support retrieval, continuation, monitoring, or background processing.
- iDialogue records and artifacts: Dialogue history, repository artifacts, Experience and Room content, business outcomes, operational records, and billing records follow their applicable iDialogue lifecycle and customer agreement.
Provider application state is not model training, public ChatGPT history, or a cross-customer data-sharing mechanism.
For a customer-provided key, the customer administers its OpenAI organization and project, including optional provider programs and available provider data controls. Provider features and contractual terms can change. See OpenAI's official Data controls in the OpenAI platform and the Data Retention Policy.
Administration checklist
Before production use:
- review provider-project access and key custody, including environment separation, rotation, and revocation;
- review optional provider programs and available data controls;
- review the agents, Skills, Salesforce fields, files, and workflows using the Connection;
- review dialogue continuity and provider application state;
- review repository artifacts and Experience or Room content;
- review Salesforce writeback and approval points; and
- monitor both iDialogue credit usage and provider-account usage.