Q&A from the Live session | Enhance Your Embedded Analytics with Generative BI

Answers generated from Webex transcript and chat interactions

Q: Why do executive summaries work on some dashboard sheets but not others in the same dashboard?
A: A few potential reasons:

  • Different sheet types: Paginated report sheets don’t support executive summaries
  • Data volume requirements: Need sufficient data/visuals to generate meaningful insights
  • Single row tables may not have enough data to generate summaries
    For specific troubleshooting, posting screenshots on the QuickSight Community is recommended.

Q: What is the licensing model for generative BI features in embedded form (executive summaries/data stories)?
A: The same licensing model applies for both embedded and non-embedded:

  • Only available with registered user embedding
  • Users need appropriate pro-level licenses (reader pro or author pro) to access generative BI features
  • Licensing requirements are the same whether embedded or not

Q: Can we embed analysis into a wiki?
A: It depends on the type of wiki:

  • Public-facing wikis require anonymous embedding, which doesn’t support generative BI features
  • Internal wikis with authenticated users can use registered user embedding to include analysis capabilities
  • For internal wikis, you can embed the console and generative BI features if using registered user embedding

Q: How accurate is the storytelling feature? Does it make up insights or only provide insights from raw data?
A: The storytelling feature:

  • Pulls insights directly from dashboard visuals and data
  • Doesn’t make up numbers or incorrect information
  • Uses LLMs for recommendations based on prompting
  • Different components include visual summaries, introductions, conclusions and recommendations
  • Accuracy varies by use case and data provided

Q: Is it recommended to use topics for creating data stories or scenarios?
A: No - topics are specifically used to power the Q&A experience. Data stories and scenarios are built directly from dashboards, visuals and provided data.

Q: How do we navigate datasets with non-natural language fields? Are there natural language keywords that are most compatible with generative functions?
A: For non-natural language fields (like abbreviated column names):

  • You can configure friendly names and synonyms when setting up topics
  • This makes datasets more natural language friendly
  • Topics allow you to define how different keywords map to your data fields