25.10 Support for Artificial Intelligence (AI) Features
Support for AI with business process (BP) workflows and records provides additional insight into the status of information in your environment.
Important: The AI features are currently available only to US-based cloud (single-tenant [ST] and multi-tenant [MT]) deployments in which the Language is set to English on the Region Format tab.
Using Artificial Intelligence (AI) Features
Variability of Summaries Generated by AI
Because of how AI large language models (LLMs) function, the summaries generated might vary, even when you generate summaries using the same input data and output options. The AI model makes billions of probabilistic choices when generating text, similar to how a human might rephrase the same idea between different conversations. This approach is by design, with the intention of making the summaries feel natural and dynamic, rather than repetitive and stale. While the wording, ordering, and formatting might vary between summaries, Oracle expects that the core information and accuracy is correct.
Translation of International Data Generated by AI
When the data submitted for summarization is in a language other than English, the model first performs a translation to derive meaning. It then generates the summary in English, leveraging its capability for natural language generation (NLG). This approach provides cohesive summaries of the information, regardless of language.
However, it is important that you are aware of the inherent limitations of machine translation, particularly when it involves technical or domain-specific information. The AI feature relies on patterns learned from vast datasets, but it can occasionally struggle with highly specialized terminology, jargon, or cultural nuances that are not well-represented in its training data. This can lead to inaccuracies or a loss of precise meaning. For this reason, Oracle does not guarantee the accuracy of the summaries when the original data contains languages other than English.
Last Published Tuesday, October 7, 2025