Tableau Next: what changes when your BI platform is an agent, not a dashboard

For twenty years, business intelligence had the same shape: someone builds a dashboard, everyone else looks at it, and the questions the dashboard doesn’t answer pile up in a queue behind the analytics team. Tableau Next is Salesforce’s bet that the shape is about to change — that instead of a person building a view and other people consuming it, an agent builds the view on demand, in response to a question asked in plain language, grounded in a governed definition of what the numbers mean.

Salesforce calls it “the world’s first agentic analytics platform,” which is marketing, but the architecture underneath is a genuine departure. Tableau Next is API-first, built directly on Data 360, and organized around a semantic model that agents query instead of guessing at raw tables. If you’ve read our piece on grounding agents on a semantic layer, you already know the punchline of why that matters: point an AI at your raw data and ask for “revenue,” and it will confidently invent a number. Tableau Next is what happens when Salesforce builds an entire analytics product around not letting that happen. This post is what the platform actually is, how the agentic skills work, where it fits against the Tableau you already run, and the licensing change nobody reads until the renewal.

Three names, one stack: Tableau Next, Tableau Semantics, Tableau Agent

Salesforce’s naming makes this harder than it needs to be, so untangle it first:

• Tableau Next is the new, API-first analytics platform. It runs on Data 360, connects natively to your governed data, and hosts the agentic experience. This is the product.
• Tableau Semantics is the semantic layer — the governed definitions of your metrics, dimensions, and relationships. It’s what makes an agent’s answer trustworthy instead of plausible. Tableau Next is built on it, and so is the grounding story for Agentforce.
• Tableau Agent is the in-flow AI assistant — the thing that turns “show me pipeline by region trending down” into an actual chart. It reached general availability in the Cloud+ Edition around Tableau Conference 2026.

The mental model: Tableau Semantics is the meaning, Tableau Next is the platform, and the agentic skills are the labor. Miss the distinction and you’ll evaluate the demo as “a nicer Tableau,” when the actual pitch is “an analytics layer your other agents can query.”

The semantic model is the whole game

Everything good about Tableau Next flows from one object: the semantic data model (SDM). An SDM has two halves. The first is a data model — which Data 360 objects you’re using and how they relate. The second is the business definitions layered on top — the field names a human would recognize, the aggregations, the calculations, and the metrics that turn columns into meaning.

Concretely, an SDM is where “revenue” stops being a guess. You define it once — its source objects, the join relationships, the aggregation, the formula — and every consumer inherits that single definition. Conceptually it reads like this:

metric: net_revenue
  label: "Net Revenue"
  description: "Recognized revenue net of refunds and credits"
  source: Order__dlm
  measure: SUM(amount__c) - SUM(refund_amount__c)
  time_dimension: close_date__c
  dimensions:
    - region__c
    - product_line__c

A Tableau Next dashboard, an analyst writing a query, and an Agentforce agent answering a question all read that definition. They get the same number because there’s only one number to get. This is the difference the semantic layer buys you, and it’s why the SDM is the artifact you actually invest in — the visualizations are cheap once the meaning is governed.

There’s a real capability jump here over classic Tableau’s relationship to the semantic layer. In classic Tableau Desktop or Cloud, the Tableau Semantics connector lets you bring in a single Data 360 object — a data model object, data lake object, or calculated insight — to start from. In Tableau Next you bring in multiple tables and define the relationships between them inside the model. The SDM is a first-class modeling surface, not a connection you point at one table. And the February 2026 release added multi-SDM querying, so an agent can reason across more than one semantic model in a single question — the difference between “answer from the sales model” and “answer from sales and support together.”

The skills: what the agents actually do

Tableau Next ships with pre-built analytics skills — the agentic labor that does the work a human analyst used to. Three matter most, and they map cleanly to three different jobs:

• Data Pro handles data preparation and modeling. It infers schema, suggests transformations, and helps build the semantic model itself — the unglamorous data-shaping work that consumes most of an analyst’s week. This is AI pointed at the setup, not just the output.
• Concierge is the natural-language front door. Ask a question in conversational language and it generates the visualization or dashboard component to answer it. Salesforce moved the Agentforce-integrated Concierge experience to GA in the February 2026 release, which is the signal that the “ask a question, get a chart” loop is meant for production, not just demos.
• Inspector is the always-on watcher. It does real-time anomaly detection and trend monitoring, surfacing alerts when a metric deviates from what’s expected — the “tell me when something breaks” job that no human does consistently because it’s boring until it isn’t.

You can also compose these into a purpose-built Analytics and Visualization Agent with its own subagents — a Data Analysis subagent and a Semantic Model Curation subagent — so the agent that answers questions and the agent that maintains the model are governed separately. That separation is deliberate and correct: curating what “revenue” means is a higher-trust operation than drawing a chart of it, and they shouldn’t run under the same permissions.

The critical thing all three skills share: they operate on the semantic model, not the raw data. Inspector isn’t watching a random column; it’s watching a defined metric. Concierge isn’t inventing an aggregation; it’s reading one. That grounding is what separates this from bolting a chatbot onto a data warehouse and hoping.

How it grounds Agentforce — and why that’s the strategic point

Here’s the part that turns Tableau Next from “a BI tool” into “a piece of your agent architecture.” Tableau Next integrates natively with Agentforce, which means the semantic model and its skills are available to your other agents. When a service agent needs to answer “what’s this customer’s lifetime value,” it doesn’t compute it — it asks the governed metric layer, and gets the same LTV your finance dashboard shows.

That’s the same argument we made in grounding agents on a semantic layer, now productized. The semantic-layer post is about the principle — agents should read governed metrics, not invent them. Tableau Next is the platform that makes the principle a product you can buy, with the modeling tools, the query engine, and the agent integration in one place. If you’re running a fleet of agents that all reference business numbers, a shared semantic layer is how you keep them from each quoting a slightly different figure and quietly eroding trust in all of them — the metric-consistency problem we keep coming back to.

Where it fits against the Tableau you already run

The question every existing Tableau customer asks: does this replace Tableau Cloud and Server? Not on day one, and probably not cleanly. Classic Tableau — Desktop, Cloud, Server — still exists, and it can now connect to Tableau Semantics models through a connector, so your existing workbooks can read the same governed definitions Tableau Next uses. The realistic near-term picture is coexistence: your pixel-perfect, heavily-formatted classic dashboards keep running, while Tableau Next is where the agentic and conversational analytics live, both drawing on one semantic layer.

Treat it as an additive capability, not a migration you have to run this quarter. The migration that does matter is the data foundation: Tableau Next sits on Data 360, so the value you get is bounded by how well-modeled and unified your data already is. An agent answering questions off a messy, un-unified dataset gives you fast wrong answers — the same failure mode we cover in Salesforce data quality for AI. Get the identity resolution and modeling right first; the agentic layer amplifies whatever sits underneath it.

The licensing shift that surprises finance

Now the part that catches people. Tableau’s pricing model changed with Tableau Next, and it’s a different shape than the seat-based Creator/Explorer/Viewer tiers long-time Tableau buyers know.

Tableau Next uses role-based licensing with two roles: Consumer (people who consume analytics content and agentic experiences) and Creator (people who also build and publish). At time of writing, Tableau Next is available standalone with Creator licensing published around $40 per user per month, and Consumer pricing quoted through sales. The notable change: this role-based model replaced a prior consumption-based approach that metered data queries, transforms, and agentic analytics against credits — so for the analytics workload itself, you’re buying seats, not watching a meter.

But watch the boundary carefully, because Data 360 underneath it is still consumption-priced. The Tableau+ bundle makes this explicit: it packages Enterprise Edition plus Tableau Next, Tableau Agent, Tableau Semantics, a Premier Success Plan, and a block of Data Cloud credits (published bundles have included figures like 250,000 credits) — with the whole thing negotiated per customer. Industry benchmarks put effective per-user costs well above the standalone sticker once the bundle and the platform underneath are counted.

Two honest takeaways for a budget you have to defend. First, the seat price is not the whole price — the Data 360 consumption underneath is a separate meter, and an agent that answers a lot of questions is running a lot of queries against it. Model both. Second, these numbers move; Salesforce has repriced this line more than once. Pull the current rate card and bundle terms directly rather than trusting any figure — including this one — secondhand, and if you’re already paying for Data 360 credits, factor the credit-optimization discipline into the total, because agentic analytics is a demand-generator for exactly the operations that consume credits fastest.

What to actually do

Tableau Next is worth taking seriously, and it’s worth not overbuying. The move that pays off is investing in the semantic model — the SDM is the durable asset, and every agent, dashboard, and analyst that reads it inherits governed, consistent numbers. Start there, on the data you’ve already unified in Data 360, and let Concierge and Inspector prove themselves against metrics you trust before you widen the surface.

Don’t treat it as a forced migration off classic Tableau — they coexist on a shared semantic layer, and your formatted dashboards aren’t going anywhere yet. Do treat the licensing as a two-layer question: role-based seats for the analytics, consumption credits for the Data 360 underneath, and a bundle price that only a current quote will tell you honestly. And remember what the whole thing is for: the strategic value isn’t prettier charts, it’s a governed metric layer your agents can query — so an SDR agent, a service agent, and a finance dashboard all cite the same number. That consistency is the quiet foundation under trustworthy multi-agent work, and it’s the reason agentic analytics is a data-architecture decision before it’s a BI one.

Understanding the basics

What is Tableau Next?

Tableau Next is Salesforce’s API-first analytics platform, built on Data 360 and positioned as an “agentic analytics platform.” Instead of centering on human-built dashboards, it’s organized around a semantic model that AI skills query — Data Pro for data prep and modeling, Concierge for natural-language questions that generate visualizations, and Inspector for real-time anomaly and trend monitoring. It integrates natively with Agentforce, so its governed metrics are available to your other agents.

How is Tableau Next different from classic Tableau?

Classic Tableau (Desktop, Cloud, Server) is a visualization platform centered on human-authored workbooks; it can now connect to Tableau Semantics models but historically starts from a single Data 360 object. Tableau Next is built on Data 360, lets you model across multiple objects inside a semantic data model, and drives the experience through agentic skills and natural language. The two coexist on a shared semantic layer rather than one being a drop-in replacement for the other.

Do I need Data 360 to use Tableau Next?

Tableau Next is built on Data 360 as its data foundation, and its semantic models reference Data 360 objects — data model objects, data lake objects, and calculated insights. That also means Data 360’s consumption-based pricing sits underneath the role-based Tableau Next seats: the analytics licensing is per-user, but the queries the agents run still consume Data 360 credits, so both layers belong in your cost model.

---

Evaluating Tableau Next and trying to separate the semantic-model investment that lasts from the licensing that changes every quarter? Talk to us — getting the Data 360 foundation and the semantic layer right so agentic analytics is grounded, not guessing, is exactly the work we do.