Tableau 2026.3: AI Dashboards, Tableau Knowledge and Forecasting

On October 23, 2026, Tableau released version 2026.3, expanding its AI-powered analytics capabilities, enterprise data governance, and business forecasting tools. The release covers Tableau Agent, Tableau Knowledge, VizQL Data Service, Tableau Next, and Tableau Pulse. Together, the changes make it easier to ask questions directly within dashboards, ground AI answers in approved business definitions, forecast time-series metrics, and maintain targets that respond to the business context.

The release also extends programmatic access to Tableau data through APIs. Companies can reuse established analytical models in internal applications and AI agents rather than rebuilding their reporting infrastructure for every use case. Tableau 2026.3 is not a single feature bundle with a uniform license or release status: some capabilities are generally available, others remain in beta, and certain functions were announced for future availability.

Tableau Agent in Dashboards: Ask Questions About Data in Natural Language

An AI assistant inside the existing dashboard

Tableau Agent in Dashboards allows business users to work with the data in an existing dashboard by asking natural-language questions. Instead of manually combining multiple filters or asking an analyst to build a separate report, an ecommerce manager can open a dashboard containing sales, orders, and marketing metrics and ask which product categories contributed most to revenue growth last month. Another question might be which regions experienced lower revenue despite an increase in order volume.

The assistant uses the measures and dimensions available in the connected data source to produce answers. Its purpose is to help users explore analytical material without requiring them to design their own visualizations. Tableau Agent operates within the user’s existing permissions and configured data sources; it does not have unrestricted access to the organization’s information.

Dashboard Overview explains what the report contains

Dashboard Overview creates a concise explanation of a dashboard’s purpose and the measures represented in its visualizations. For example, a manager opening a marketing-performance report with advertising spend, revenue, orders, and channel efficiency can ask the assistant for an overview before exploring individual charts. This is useful in organizations with dozens of dashboards built by different teams over several years, especially for employees who do not yet know the reporting structure.

The generated overview does not replace a written methodology. If a workbook uses custom definitions of Revenue, Conversion Rate, or Customer Acquisition Cost, the company still needs to document the calculation rules. A summary of what a chart shows is different from proof that the underlying metrics are defined consistently.

Dashboard Insights identifies trends, contributors, and unusual results

Dashboard Insights examines the visualizations already present in a report and identifies patterns worth further analysis. The supported insight categories include:

• Current Trend: the recent direction of a metric.
• Trend Change: a meaningful shift in the direction or pace of a trend.
• Top Drivers and Top Detractors: segments contributing positively or negatively to a result.
• Concentrated Contribution: situations in which a small number of values account for a large share of performance.
• Unexpected Values: observations outside the expected range.
• Bottom Contributors: the lowest-performing segments.

Consider an online store whose total revenue increased 8%. Insights may show that nearly all of the improvement came from one category while several others declined. In advertising analytics, overall ROAS may look stable even though individual channels experienced substantial changes. These findings help management move beyond a single top-line number and inspect the structure of performance.

Detected relationships are not proof of causation. If advertising performance falls at the same time as product availability changes, the assistant may identify that the events coincide, but the dashboard alone cannot establish which event caused the other. Further analytical validation is required.

Dashboard Q&A explores underlying data beyond the visible charts

Dashboard Q&A lets the user ask additional questions about the data behind a dashboard, rather than only examining the charts already displayed. Examples include “Show the ten products with the largest revenue decline” and “Which paid channels generated the highest Average Order Value among new customers?” The assistant can use accessible fields in the workbook’s published or embedded data sources to answer these questions.

Where supported, results can include text, tables, bar charts, and scatter plots. This expands the questions that can be answered without changing the dashboard itself. A developer does not need to anticipate every possible breakdown in advance, provided the necessary fields exist and the user has permission to query them.

These capabilities predate the October release

Dashboard Overview and Dashboard Insights began appearing in Tableau Cloud earlier in 2026, and Dashboard Q&A also entered beta before version 2026.3. The October release should therefore be understood as an expansion in capability, integrations, and availability rather than the first introduction of all three functions. Their behavior can also differ by product, edition, and deployment type.

How to Enable Tableau Agent in Dashboards

Licenses, site settings, and workbook permissions

At the time of the 2026.3 release, Tableau Agent in Dashboards was in beta in supported Tableau Cloud+, Tableau+, and Tableau Server+ configurations. Tableau Server requires version 2026.3 or later and an explicit beta opt-in. Administrators must enable AI in Tableau, and users must have the relevant workbook permissions, including AI Access and Full Data Query.

1. Verify that the organization has a supported Tableau edition and deployment.
2. Confirm that the site administrator has enabled AI in Tableau.
3. Enable Tableau Agent for Dashboard Consumers where the feature is supported.
4. Check that the intended users can open the workbook.
5. Grant and verify AI Access and Full Data Query permissions.
6. Open a supported dashboard and enable the appropriate Tableau Agent features.
7. Test Dashboard Overview, Dashboard Insights, and Dashboard Q&A under actual user accounts.

Dashboard authors can control the availability of individual Agent features for a report. A company might permit overview and insight generation but restrict ad hoc questions against underlying data. Permission settings should reflect the organization’s actual data-governance policy, not merely whether a feature is available in the interface.

Row-level security still applies

Tableau Agent honors existing authorization mechanisms, including row-level security. A regional sales manager who can see only one territory should not receive results for other regions just because the request is written as a question instead of built with dashboard filters. Teams should verify that behavior using accounts with different permissions before wider adoption, particularly in multinational or multi-brand organizations.

Tableau Knowledge: Governed Business Context for AI Analytics

Knowledge Graph connects data assets, documents, and definitions

Tableau Knowledge introduces an enterprise knowledge-management layer intended to help AI agents interpret a company’s data according to its own business rules. It connects analytical assets and business context in a Knowledge Graph. The relationships help the assistant understand not only which fields exist, but how those fields should be interpreted.

A retailer may combine Shopify orders, GA4 behavioral data, Salesforce CRM records, Google Ads and Meta Ads campaign results, and Tableau reports. Each system can use a different definition of Revenue. Shopify may reflect an order value after certain adjustments, GA4 may report ecommerce event revenue, and advertising platforms may show attributed conversion value. Without context, an assistant could mistakenly treat these values as interchangeable.

Tableau Knowledge helps standardize the meaning of a metric across reports and AI queries. The goal is to document the relationship between analytical objects and business definitions, reducing the need for the model to infer a calculation rule on its own.

Objects included in the Knowledge Graph

The graph can connect published data sources, workbooks, dashboards, sheets, and embedded data sources, along with Business Preferences that express the organization’s additional definitions and interpretation rules. For example, the Net Revenue metric can be connected to a particular order data source and an Ecommerce Performance dashboard, with supporting rules explaining whether refunds and canceled orders are included.

These relationships provide a more consistent context when an AI agent answers questions across different analytical assets. They also make it easier for the data team to inspect which reports use a particular field or source and where documentation may be missing.

Business Preferences specify calculation rules in plain language

Business Preferences lets organizations tell AI systems which definitions to use. An ecommerce team might specify “Exclude canceled orders and the value of returns when calculating Net Revenue,” or “A New Customer is a buyer with no previous paid orders in the order system.” A performance-marketing team might direct the assistant to use actual store revenue rather than attributed ad-platform conversion value when measuring overall commercial results.

These instructions matter in businesses with nonstandard definitions of gross margin, customer acquisition cost, or retention. A B2B SaaS company may distinguish New ARR, Expansion ARR, and Renewal ARR, while a service company may count qualified inquiries separately from raw form submissions. Without governance, an assistant can select the wrong field or apply an inappropriate aggregation.

Global preferences and workbook-specific preferences

Tableau distinguishes between Global Business Preferences stored in Tableau Knowledge and Workbook Business Preferences attached to individual workbooks. A company can define Revenue globally while maintaining special calculation rules for a finance dashboard. In version 2026.3, Tableau Agent in Dashboards can use global preferences from Tableau Knowledge alongside the context defined for an individual workbook.

Local workbook instructions may narrow or override the general context, so reporting teams should review conflicts. Centralizing rules does not eliminate the need for consistent ownership and version control of metric definitions.

How to Configure Tableau Knowledge

Enable the feature and create a graph

Tableau Knowledge is a Tableau Cloud capability, available at general availability in Tableau Cloud+ and Tableau+. An administrator must enable it under Tableau Cloud → Settings → AI Settings by Feature and ensure the necessary Salesforce organization configuration is in place. A feature being included with an edition does not mean it is automatically switched on.

After activation, the team can create a Knowledge Graph, select the relevant analytical assets, add existing data sources and workbooks, inspect automatically discovered relationships, and add business definitions. The graph can then be reviewed and set as the default context for applicable AI queries. Graph View shows relationships between objects; Explorer View presents the structure hierarchically.

1. Choose the Tableau analytical objects that belong in the graph.
2. Add the available data sources and workbooks.
3. Inspect relationships detected between the objects.
4. Add and review metric definitions and Business Preferences.
5. Upload supporting documents containing approved business rules.
6. Review the graph structure in Graph View and Explorer View.
7. Configure the appropriate default Knowledge Graph.
8. Test Tableau Agent responses against known, independently validated answers.

Upload documented definitions instead of improvising them

Business Preferences can be added manually or imported from company documents. The supported document types described for this workflow include PDF, DOCX, TXT, Markdown, and CSV. A marketing team could provide a glossary covering Revenue, Net Revenue, Average Order Value, Customer Acquisition Cost, Customer Lifetime Value, Marketing Qualified Lead, Sales Qualified Lead, New Customer, and Returning Customer, each linked to an authoritative source and calculation method.

The definitions must be consistent. If different documents contain incompatible rules, an AI assistant cannot be expected to resolve those conflicts reliably. Reviewing the measurement methodology before uploading documents is therefore part of the implementation process, not optional postlaunch cleanup.

Quality Scorecard and availability limitations

Knowledge Graph Quality Scorecard helps identify weaknesses such as undocumented fields, ownerless sources, unverified analytical objects, disconnected assets, or insufficient business context. For example, two similarly named Revenue fields without descriptions can lead to the wrong selection in an AI query. Improving their definitions and relationships makes the result easier to audit.

At release, Tableau Knowledge was generally available in Tableau Cloud+ and Tableau+, while Tableau Server+ support was planned for a later period. Its initial language support was English. International organizations should account for that limitation when preparing metadata and company documentation.

VizQL Data Service: Access Tableau Data Through an API

Reuse analytical sources without creating new dashboards

VizQL Data Service (VDS) provides programmatic access to Tableau data through HTTP APIs. Instead of opening a workbook and interacting with a visualization, a developer can send a structured query to a supported data source and receive the underlying results. This is useful when a company wants to reuse its Tableau analytics in an internal marketing portal, executive application, chat assistant, automated KPI monitor, or AI agent.

Consider an organization that combines Google Ads, Meta Ads, GA4, and Shopify data into a reporting model in Tableau. An internal tool can retrieve relevant data through VDS without requiring a separate manually maintained set of calculations for every application. The value is the reuse of governed analytical logic and access controls, not unrestricted access to all enterprise databases.

Workbook data sources, interactive sessions, and metadata

The service has expanded to support applicable workbook data sources, including embedded sources and elements of the workbook’s analytical logic. Developers may be able to query calculated fields, groups, or other supported characteristics of the data model without publishing a duplicate data source solely for another application.

For embedded analytics, a query can also be tied to an active user session. If a user selects a particular region and date range inside an embedded dashboard, the application can use session context when retrieving data for a nearby AI assistant. An example question would be “Explain the sales change in the selected region.” The implementation must preserve authentication, session integrity, and row-level access rules.

VDS also exposes metadata about the fields and other supported characteristics of the source. Applications and AI agents can inspect available measures and dimensions before assembling a query. Developers must still respect authorization, API limitations, and compatibility with the installed Tableau version.

Availability and release history

VDS is generally available in Tableau Cloud and Tableau Server, but it was introduced before the 2026.3 release. October’s changes continue the expansion of query and interactive-session capabilities rather than introducing the base service for the first time. Tableau Server deployments require attention to the installed server version and the particular API feature being used.

Tableau Next: Moirai Forecasting for Business Metrics

Forecast time series from historical observations

Moirai Forecasting became generally available in Tableau Next. It uses an AI time-series foundation model to forecast future values from historical data. Rather than configuring a separate traditional statistical model for every individual series, the user can choose a time field and numerical measure in the analytical interface and generate a forecast using available observations.

An online store might forecast daily revenue or order volume, while a service business might estimate future lead or support-ticket demand. A subscription business may use the same capability to examine recurring performance patterns. The resulting forecast is only as useful as the historical series, its coverage, and the assumptions about future business conditions.

How a marketer might use Moirai Forecasting

A practical workflow starts by selecting the approved semantic model or time-series dataset, choosing the measure to forecast and the relevant date field, and defining the intended horizon. The user then reviews the generated forecast and compares it with historical patterns. Seasonal demand, promotions, inventory availability, and structural breaks all need consideration when deciding whether the result is credible.

For example, an ecommerce team could use past sales to prepare an order forecast before purchasing inventory. The forecast may help define an expected range for the coming weeks, but it should not be treated as an automatically reliable estimate of the next campaign’s incremental sales.

Forecasting seasonal ecommerce demand

Demand is influenced by more than a calendar sequence. Black Friday, promotional campaigns, stockouts, category launches, and pricing changes may produce patterns that do not repeat from one year to the next. Analysts should compare AI forecasts with merchandising plans, current stock levels, paid-media schedules, and any extraordinary events missing from the historical data.

A forecast based on last year’s order series may describe the historical trajectory accurately while still failing after a major assortment change. Teams should monitor prediction errors over time and reassess the model when conditions shift.

Limitations of AI-based forecasting

Moirai Forecasting estimates a future value; it does not establish the causal effect of marketing spend. The result depends on data quality, the amount of history, and whether the historical pattern remains relevant. A sudden market disruption, major campaign change, or previously unseen product cannot be modeled reliably from an unchanged past trend alone. Forecasts should be reviewed alongside business judgment and, where appropriate, separate causal or experimental analysis.

Dynamic Goals: KPI Targets That Change With Business Context

Targets are linked to a governed semantic model

Dynamic Goals, generally available in Tableau Next, addresses the maintenance of targets across many business segments. Instead of manually entering and updating individual targets for every region, store, product, or territory, a company can link the target to a measure defined in its Semantic Model. Targets then adjust based on the selected filters, period, and relevant business context.

A multinational retailer may maintain revenue plans by country and product category. When a manager opens a dashboard for Germany, the target can reflect Germany’s plan. Switching to France brings up the corresponding value rather than leaving a single universal threshold on screen. The same principle can apply to store-level targets or regional sales teams.

Example: a regional retail network

Suppose a retail chain sets different monthly sales goals for each store and region. A regional manager should see both the actual revenue and the target applicable to the selected territory. With Dynamic Goals, Tableau Next can retrieve the target from the governed model instead of requiring analysts to configure separate static goal cards for each location.

The feature is particularly relevant where targets change by time period or business segment. Accuracy still depends on the completeness of the target data and the consistency of the semantic model, including the mapping between metrics and dimensions.

Configuring and validating Dynamic Goals

The implementation requires the team to prepare the metric and its goal data in the semantic model, map the relevant dimensions, and verify that selected filters resolve to the intended target. Test different combinations of regions, products, and reporting periods. A dynamic calculation should not silently display a misleading target when the underlying goal is missing or incompatible with the selected filters.

Tableau Pulse: Data-Based Actions Connect KPIs to Records

Contextual links from a metric to another system

Data-based Actions, generally available in Tableau Cloud, extends Tableau Pulse by letting analysts attach contextual links to the dimensions contributing to a metric. A sales manager examining new deals can move from a KPI or segment breakdown to the relevant CRM records. An ecommerce manager who notices weaker category sales can follow a link to the applicable product or category information in the commerce system.

This turns a metric into a starting point for an operational workflow, reducing the need to search for the same entity again in another application. It does not replace the destination system or automatically perform the action on the user’s behalf.

Dimension values can personalize destinations

Links can incorporate dimension values as variables so that the destination corresponds to the selected item. For example, a store identifier or territory may be passed in a URL to open the appropriate view in another system. Teams should test URL construction and access permissions rather than assuming any linked record will be available to every viewer.

Data-based Actions does not edit records by itself

Data-based Actions is a contextual navigation mechanism, not a general-purpose write-back API. Opening a CRM record does not mean that Tableau Pulse has changed its status, created a task, or updated customer data. Any subsequent operation remains subject to the destination application’s permissions and workflow controls.

Other Tableau 2026.3 Updates

Tableau Data Apps: AI-assisted analytical applications

The release also included Tableau Data Apps in beta, supporting the development of custom applications based on Tableau data. A company might build an internal interface that combines governed metrics, filters, and operational links for specific teams. Compatible AI development tools can assist with application creation, while the organization continues to rely on Tableau for data connectivity, security, and managed analytical resources.

Because Tableau Data Apps was still in beta, companies should treat it as an emerging capability rather than a fully mature replacement for conventional internal application development.

Tableau Agent Proactive Intelligence: future monitoring

Tableau Next also previewed Tableau Agent Proactive Intelligence as a coming-soon capability for surfacing changes, risks, and emerging patterns without requiring a user to ask a question first. This is distinct from features that were already generally available in 2026.3. The announcement should not be interpreted as proof that all proactive workflows were production-ready on October 23.

Transparent Tooltips

Tableau Desktop, Tableau Cloud, and Tableau Server gained Transparent Tooltips, a presentation improvement intended to reduce visual obstruction when exploring charts. Although this is less significant than the AI and data-governance features, it can help dashboard consumers keep nearby data in view when inspecting a tooltip.

What the Release Means for Performance Marketing

Performance-marketing teams often combine Google Ads, Meta Ads, TikTok Ads, GA4, ecommerce orders, and customer data. A Tableau report might include Advertising Spend, Orders, Revenue, Customer Acquisition Cost, ROAS, Conversion Rate, Average Order Value, New Customers, and Returning Customers. Tableau Agent can help users explore that model through questions such as “Which paid channels increased spending faster than revenue?” or “Which categories experienced lower conversion rates while traffic remained stable?”

Tableau Knowledge can clarify that actual revenue from the commerce platform is not necessarily the same as the conversion value attributed by each ad network. With approved definitions, teams can align Revenue, New Customer, and ROAS across reports. VizQL Data Service can also make the governed metrics available to internal performance dashboards and automated monitoring systems.

The analytical benefit depends on the underlying conversion tracking, attribution, and data integration. AI cannot reliably reconcile inconsistent order counts or campaign taxonomies that the organization has not resolved. Attribution models and experimental measurement remain separate analytical disciplines.

What the Release Means for Ecommerce

Sales and assortment analysis

Ecommerce teams can use Tableau Agent to investigate changes in sales by category, product, and region, including the segments responsible for most growth or decline. Follow-up questions can reveal patterns in the underlying dataset without constructing another workbook for every investigation.

Demand forecasting and planning

Moirai Forecasting in Tableau Next can project future order or revenue trends from historical data. Merchandising teams may use the resulting estimates when planning purchases and inventory, while accounting for promotions, seasonality, assortment changes, and stock constraints.

Commercial targets and operational follow-up

Dynamic Goals supports targets that vary by country, store, or category. Pulse Data-based Actions can then help users move from a problematic KPI to the associated operational records. These capabilities serve different steps in the workflow: forecasting possible outcomes, comparing actuals with goals, and navigating to details that require action.

What the Release Means for Customer Services

Service businesses can use Tableau to monitor inquiries, Average Response Time, Average Resolution Time, Customer Satisfaction, repeat contacts, employee workload, and cost to serve. Tableau Agent can help identify which categories contributed to a recent increase in resolution time. Moirai Forecasting may estimate incoming contact volume based on historical patterns, while Dynamic Goals can display distinct service-level targets by department or location.

AI-generated explanations still require operational verification. A longer response time might coincide with higher demand, a staffing issue, or a change in case complexity. The assistant can surface relevant segments and patterns, but it does not establish the root cause without additional evidence.

What the Release Means for B2B SaaS

B2B SaaS companies can combine Salesforce CRM with product analytics to monitor Marketing Qualified Leads, Sales Qualified Leads, Pipeline Value, stage conversion, Annual Recurring Revenue, Customer Churn, Net Revenue Retention, and Customer Lifetime Value. A manager might ask which industries created the most new pipeline last quarter or which customer segments have the highest churn.

Tableau Knowledge is relevant because SaaS businesses often apply specific recognition rules to subscriptions, expansions, renewals, and cancellations. Tableau Agent can explore approved data, while Dynamic Goals can connect targets to business units, regions, or customer segments. Reliable definitions are necessary before comparing performance across periods or teams.

Implementation: Prepare the Measurement Foundation First

1. Audit source systems and integrations

Identify the authoritative systems behind each core measure: advertising platforms for campaign spending and ad results; GA4 for user behavior; Shopify or another commerce platform for orders; Salesforce for CRM and sales; and billing systems for payments and subscriptions. Verify that events, identifiers, and integrations are sufficiently complete. Tableau cannot reconstruct events that were never captured by their source systems.

2. Agree on KPI definitions

Document Revenue, Conversion Rate, ROAS, CAC, LTV, and the other metrics that may vary between platforms or teams. For each definition, specify the source, calculation, exclusions, attribution assumptions, and responsible owner. These definitions provide the basis for trustworthy Business Preferences and Knowledge Graph relationships.

3. Review data permissions

Verify workbook access, AI permissions, and row-level security. Run the same natural-language questions under accounts with different roles to confirm that the assistant respects the authorized scope. Data-governance testing needs to cover both visible dashboards and underlying-data queries.

4. Validate AI answers against known reports

Prepare benchmark questions with independently validated answers. Compare the assistant’s results with the original reporting system, including filters, aggregation rules, and period selection. Investigate disagreements before allowing teams to rely on AI explanations in operational decisions.

5. Test forecasting on historical periods

Assess forecast accuracy by training on earlier periods and comparing estimates with subsequent actual outcomes. Inspect error by season, product category, region, and business events. Avoid treating a plausible-looking forecast chart as a substitute for measured predictive performance.

6. Connect monitoring to safe operational actions

Configure who receives alerts or sees the relevant KPIs and how users move from analytical findings to operating systems. Validate links, permissions, and escalation rules. Document the distinction between a detected anomaly, a recommendation, and an approved change to business processes.

Availability of Tableau 2026.3 Features

The Tableau 2026.3 release spans multiple products and editions. The following statuses describe the capabilities at the time of the October 2026 announcement, not a guarantee of access for every customer:

Installing or accessing Tableau 2026.3 does not automatically grant every capability. Administrators should verify the edition, deployment version, AI feature activation, license terms, and beta enrollment for each function. Actual availability can differ between Tableau Cloud, Tableau Server, and Tableau Next.

Tableau 2026.3 and the Future of Enterprise Analytics

Tableau is a Salesforce analytics platform used to combine enterprise data, visualize performance, and support business decisions. Version 2026.3 extends the product in four broad directions. First, conversational analytics allows users to explore dashboard data with natural-language questions. Second, Tableau Knowledge introduces a governed context for interpreting data and business rules. Third, Tableau Next expands forecasting and dynamic target management through Moirai and Dynamic Goals. Fourth, VizQL Data Service and Pulse Data-based Actions help connect analytical assets with applications and business workflows.

The release does not remove the need for analysts, data governance, or reliable source systems. Its practical value depends on whether an organization can make existing measurements consistent, explainable, appropriately permissioned, and accessible to the teams that need them. For performance marketing and ecommerce, the strongest implementation path starts with conversion and commercial data quality, not with the AI interface itself.