High-Level Architecture

The following is the high-level architecture for Generative AI (GenAI) in Reports:


high-level architecture

  1. A user launches a report in Narrative Reporting that either has "Conditional Text" defined with GenAI based AutoTextSummary function or "Notes Summarization" enabled.

  2. In the background, data is retrieved from the report’s data source.

  3. For the GenAI feature:

    • Generating Narrative Summaries: The report grid data that satisfies the pre-defined conditions, as well as the prompt from the corresponding Narrative Reporting machine learning property file, are sent to the Fusion AI service.

    • Summarizing Notes: The report grid data (if Use Context is enabled) and the Notes, as well as the prompt from the corresponding Narrative Reporting machine learning property file, are sent to the Fusion AI service.

  4. The Fusion AI service generates a response and sends it back to the Narrative Reporting instance.

  5. The report is augmented with the generated response and is rendered to the user.

Note:

Due to the additional processing required and network traffic to the Fusion AI service, which exists on a different instance than Narrative Reporting, reports will take additional processing time when utilizing the GenAI feature. If the conditional text is applied to a larger number of cells in the report, and/or if GenAI reports are included in Books or Bursting and run for multiple POVs, this could negatively impact report processing time. Thorough testing should be performed to ensure your process still meet any service level objectives you have for running reports for your users.