Tableau Next displays a summary of the insights related to those metrics. Tableau Next
creates the insight summary differently, depending on whether Tableau Agent with conversational
analytics is enabled or not. When the agent isn't enabled, insight summaries are created based
on the metrics you follow (two or more required). When the agent is enabled, Tableau Next
creates insight summaries using Agentforce (no followed metrics needed).
Required Editions
| User Permissions Needed |
| To explore metrics and insights: |
Tableau Unmetered Consumer or Tableau Next Consumer or Tableau Unmetered Self-Service Consumer or Tableau Next Self-Service Consumer
permission set |
When Tableau Agent with conversational analytics is enabled in Tableau Next, metric insight
summaries are automatically generated in your preferred language and don't require followed
metrics. These insight summaries are available in the metric details page and for the top
ranked insight in a metric card.
When the agent is turned off, Tableau Next leverages a large language model (LLM) to
summarize the most significant changes, considering Period Over Period Change or Unusual
Change insights detected by the system. The system prioritizes summarizing metrics with
higher-impact insights (for example, metrics with unusual change are ranked higher).
The resulting insight summary provides a personalized overview in plain, natural language
with rich text reference on the metric names to be found at the top of the Following page. To
maximize consistency and performance, insight summaries are generated up to once every 24
hours upon your last homepage visit and cached for subsequent visits within the day.
The insights used for summaries are generated using natural language templates with values
calculated using deterministic statistics. The prompt used to summarize these factual insights
is thoroughly tested and consistently iterated.
Metric insight summaries are built on the Einstein Trust Layer, enabling trusted, ethical, and
open AI-powered experiences without compromising data security and privacy. Like all
generative AI features in Salesforce, users can also provide optional feedback on output they
find to be biased, inaccurate, or inappropriate to help to improve the experience.
