AI in Tableau Usage

AI in Tableau Cloud is powered by Einstein Generative AI. If you use generative AI features in Tableau Cloud, for example, asking Tableau Agent to help build a viz or suggest a description for a data source in Tableau Catalog, it consumes Einstein Requests and possibly Data Cloud credits (Data services credits).

Einstein Requests are a consumption metric for generative AI and are consumed from your total Einstein Requests allotment whenever you use an AI in Tableau feature. For a list of AI in Tableau features, see AI in Tableau Features(Link opens in a new window).

Data Cloud credits are consumed by the generative AI audit data (also known as audit trail) feature, which enables you to track the usage of generative AI in your Tableau sites and Salesforce org.

The number of Einstein Request credits consumed varies based on the number of prompts needed and the length of the prompt (both system and user) sent to our Large Language Model (LLM) providers. The number of Data Cloud credits consumed varies based on the number of rows or records that are processed.

Note: Insight Summaries (part of Tableau Pulse) is an AI in Tableau feature. It doesn’t consume Einstein Requests and is available in all editions of Tableau (Standard, Enterprise, and Tableau+).

Flow diagram showing how Tableau Agent consumes Einstein Requests and Data Cloud credits.

Einstein Requests

Direct calls to a supported Large Language Model (LLM) gateway impact your credit consumption. The number of Einstein Requests consumed varies and can differ by feature and how the user interacts with the feature.

You cannot allocate a portion of your credits by feature. For example, if you use Tableau Agent for viz authoring, for cleaning data in Tableau Prep, or for Tableau Pulse Enhanced Q&A (Discover), Einstein Requests are consumed from the overall pool of credits that are allotted to your org.

For more information about the usage rate for Einstein Requests, see the rate card for Einstein Requests(Link opens in a new window).

The Einstein Request consumption calculation is based on:

  • The number of LLM API calls needed for each feature use.

  • The length of the prompt sent to the LLM providers.

  • The response returned (measured in words) from the LLM for each LLM API call.

When conversing with Tableau Agent, be specific about what you want Tableau Agent to do and limit unnecessary back and forth conversation. This can help conserve Einstein Request usage, while still returning the best results. For example if you type “Hi Tableau Agent”, or “Thanks, this is just what I needed!”, this will trigger a call to the LLM. For more tips on how to get the most out of Tableau Agent, see Tips for getting the best results from Tableau Agent(Link opens in a new window).

LLM API calls needed per feature use

The number of LLM API calls required per feature use depends on the complexity of the task that the feature performs. For example, for Tableau Agent to create a visualization requires multiple LLM API calls:

  • One LLM API call to interpret the query and map it to a supported skill set.

  • One or more LLM API calls to perform the task of creating the visualization.

Call size

Einstein Request consumption is also impacted by the size of the call made to the LLM. This is a combination of the length of the prompt (number of words) sent to the LLM as well as the length of the response received.

Call size is determined by these factors:

  • The number of words in the system (base prompt) developed by the Tableau engineers. This can also include the number of words injected into the prompt from the data source ( metadata, insights, etc) to ground it for accuracy.

  • The number of words in the user's request to the LLM.

  • The number of words in the LLM response.

Calculating Einstein Request Consumption

The Einstein Request consumption usage calculator for AI in Tableau features includes:

  • A usage type multiplier associated with the LLM used.

  • An API call size factor associated with the size of the API call.

The API call size factor, Einstein Request multiplier, and calculation examples can be found in the rate card for Einstein Requests(Link opens in a new window). AI in Tableau currently supports the Salesforce-enabled foundational LLMs only, so the Standard Einstein Generative AI usage type applies.

Data Cloud credits

Data Cloud credits aren't directly consumed by using AI in Tableau features. Instead, these credits are consumed by the Einstein Trust Layer, that AI in Tableau is built on top of, through a generative AI audit data feature known as Audit Trail.

Audit Trail lets your information security teams monitor the Einstein Trust Layer features, including pattern-based data masking and toxicity detection. This feature uses Data Cloud credits (allocated during the purchase of AI in Tableau), for ingestion, storage, and processing AI usage data.

Audit trail helps maintain the safety and accuracy of the generated AI responses. It stores the prompt and response for every LLM API call along with other audit data for a maximum of 30 days in your Salesforce Data Cloud instance. For more information about the types of data collected by the Generative AI Audit and Feedback feature, see Generative AI Audit and Feedback Data(Link opens in a new window) in Salesforce Help.

Data Cloud credits are consumed when Audit Trail ingests data during an LLM API call. 1 LLM API call ingests 24 rows of Audit Trail data into Data Cloud and uses the Batch Data Pipeline usage type.

Note: Data Cloud Credit consumption also comes from a pool of allotted credits. If you generate an Einstein Generative AI Audit and Feedback Data report, for example, it uses the Data Queries usage type and consumes Data Cloud credits. If you have other Salesforce Cloud features that consume Data Cloud credits, then this can also impact the available credits in your pool. For more information, see Data Cloud Billable Usage Types(Link opens in a new window) and Access Einstein Generative AI Audit and Feedback Data Reports and Dashboards(Link opens in a new window) in Salesforce Help.

To estimate the number of Data Cloud credits, see Data Cloud Multiplers(Link opens in a new window). For more information about usage types for audit and feedback, see Billing Considerations for Audit and Feedback(Link opens in a new window) in Salesforce Help.

You can also use Tableau to query Data Cloud and build visualizations of your AI audit and feedback data. For more information, see View Einstein Request Usage.

Billing considerations for AI in Tableau

AI in Tableau uses the following billable usage types. To help you keep track of your usage, you can use Digital Wallet, a free account management tool in Salesforce Data Cloud that offers near real-time consumption data for enabled products. Digital Wallet shows you the total number of Einstein Requests and Data Cloud credits consumed as well as your remaining balance.

Digital Wallet also offers pre-built dashboards showing the number of users, number of requests, and other details. You can also connect to the Digital Wallet tables in Data Cloud from Tableau and build visualizations for your consumption data. For more information, see View Einstein Request Usage.

Digital Wallet Card Usage Type Usage Type Description Notes
Einstein Requests Standard Einstein Requests Usage is calculated based on the number of calls to the LLM gateway if the gateway uses a Salesforce LLM. To learn more, see the Rate Card for Einstein Requests(Link opens in a new window)

All AI in Tableau features use Salesforce-enabled foundational LLMs, so the Standard usage type applies.

Einstein Requests are consumed when you use AI in Tableau features. This includes Tableau Pulse Enhanced Q&A (Discover) features, Tableau Agent in Tableau Prep (Tableau Builder and Web Authoring) and Viz authoring (Tableau Desktop and Tableau Cloud), and Tableau Catalog.

For more information about AI in Tableau features, see AI in Tableau Features.

Data Services Batch Data Pipeline

Usage is calculated based on the number of rows batch data processed by Data Cloud data streams across all connectors, with the exception of structured data ingested via the Internal Data Pipeline.

Pipeline.

To learn more, see Billing Considerations for Audit and Feedback(Link opens in a new window) in Salesforce Help.

Audit and feedback data are ingested into Data Cloud data streams, and the usage is based on the amount of data ingested.

On average, each round trip to the large language model (LLM) and back results in 24 records being ingested into Data Cloud.

The volume of data ingested is the primary contributor to the consumption of credits among the three types of usage.