Tableau Data Apps are task-specific applications built on Tableau's governed knowledge, generated from a plain-language description and able to run in Slack, Teams, Salesforce, Claude or ChatGPT. With them, Tableau stops being only a BI platform and becomes an app platform. The dashboard is one of the things it makes.
I have hired, trained and certified more Tableau developers than I can count. Some of the best dashboard craftsmen in Europe have worked for Biztory.
So the last two glossary terms – data apps and dashboards – are the ones I have the least distance from. I want to be honest about that before I explain them, and honest about the fact that we started acting on them long before Dreamforce made them official.
What are Tableau Data Apps?
Tableau Data Apps are purpose-built, interactive applications that run on Tableau's governed data and knowledge. Each one is tailored to a single task, shows only what that task needs, and usually ends in an action.
They can be built by describing what you want in Tableau Studio or in AI tools such as ChatGPT, and then hosted and governed in Tableau.
The glossary's key phrase is "a specific task." That is what separates a data app from a dashboard.
Dashboards:
- Built around: A subject: sales, inventory, service
- Purpose: Explore, compare and form a judgement
- Shows: A broad view for many questions
- Ends in: A human decision, taken elsewhere
- Typical user: Analysts and managers who explore
Data app:
- Built around: One job; approve this discount, release this blocked order, etc.
- Purpose: Complete a task and act on it
- Shows: Only what the job needs, with the governed reasoning behind each number
- Ends in: An action, often in the same screen
- Typical user: The person doing the job, often not a BI user
The dashboard stays as the place people think. The data app arrives as the place people act. Both read from the same knowledge graph, and a good analytics team will ship both.
Data apps: the dashboard gets a sibling
Built around one job, not one subject
A dashboard is a general-purpose view of a subject. It exists so a human can explore, compare and form a judgement, and nothing in this glossary replaces that.
A data app is built around one job: decide whether this warranty claim is covered. It shows only what the job needs, includes the governed reasoning behind the number, and ends in an action. Tableau Next has been moving this way for a year.
Dreamforce made it explicit.
Build once, run anywhere
What makes Data Apps a platform feature rather than a nice demo is where they run. Salesforce was specific: the apps extend into Slack, Teams, Salesforce itself, Claude and ChatGPT, through the Tableau MCP servers already live for Next, Cloud and Server. TechTarget's Dreamforce coverage describes Data Apps as extending Studio's app-building capabilities to third-party environments such as ChatGPT and Claude.
A warranty desk built once can be opened by a workshop manager in Teams and by a field agent asking Claude, with the same governed logic behind both. That is a distribution model no analytics vendor has had before.
What we shipped before Studio
We were building this way already before Tableau Studio was announced. Our first Tableau Next app on AgentExchange, shipped for the launch, takes a prompt and produces a governed model and dashboard inside the customer's own Salesforce org. No data leaves. No hand-built first draft.
When the product team at Tableau told us we were the first partner that clearly understood Salesforce, I took it as a sign we had read the direction right, not that we had guessed.
Governed vibe coding: Tableau Studio builds applications, not charts
Vibe coding means building data apps and workflows by describing what you want in your own words, and letting AI turn that intent into a working solution without writing code. Tableau calls its version governed vibe coding, and the adjective carries the whole claim.
From description to working app
Tableau Studio, the successor to Desktop with general availability expected by the end of October, builds a working application from a description. Our Tableau Studio explainer covers how the workspace itself works.
Not a mock-up and not only a dashboard. A connected, permissioned, multi-step app on the knowledge graph: the screen, the logic, the governed query behind each number, and the action at the end. I watched it at Dreamforce and found no reason to doubt it.
Why "governed" is the whole claim
This is where I will give Tableau the strongest claim in the whole series. Plenty of tools now generate a chart from a prompt, and a few generate a passable app. Studio generates a governed one:
- The metrics are the signed-off metrics from Tableau Knowledge
- The permissions are the real permissions
- The actions land in the real system of record through the Decision Engine
- The whole thing is visible in the Command Center alongside every other agent in the company
Generating a pretty screen is easy. Generating a working application you can hand to a warehouse manager without a review cycle is the bar. Studio is the first product that credibly sets it.
The risk: app sprawl
One honest caveat; Studio and Data Apps could lead to complications. When anyone can build an app in minutes, you get piles of near duplicates.
The Command Center exists partly for that reason. Governance of who builds what, on which definitions, is now part of the analytics job. Ignore it and the platform will drown you in apps the way the last one drowned you in workbooks.
What our own tests showed
We had tested the same idea with general-purpose models before Studio shipped. Give a capable model the knowledge graph and a good description of the task, and it produces a first version at least as good as what a mid-level developer builds in a week.
Sometimes better, because it has no habits. Usually worse on the last ten percent, because taste is not in the semantic model.
So I will say this plainly: the first draft of a dashboard is no longer where the value sits. A great builder still turns a correct draft into one people trust and return to, and I have watched our best people do exactly that with a generated starting point.
The craft moves up. Part of it we encode as best practices the model follows. The rest stays with the person who knows why a chart is wrong even when the numbers are right.
What survives when the first draft is free
I put the question to the whole team well before Dreamforce. The consensus came back as one sentence: the building, yes, but not the thinking. I would sharpen it. What survives is accountability, encoded judgement and the dashboard itself.
Accountability
Someone signs off that the number the app shows is the number the business acts on. A model does not carry that. A person does, and needs a method to carry it well. That is what our Blueprint is for, and why it matters more now, not less.
Encoded judgement
Every exception, every business rule, every "yes, but not in week 52" that a good analyst carries in their head has to be written down where the agent and the app builder read it.
That work is slow, needs real domain knowledge, and compounds. It is the asset that makes the next vibe-coded app correct instead of merely fast. Part 2 of this series covers how Tableau Knowledge gives it a home.
The dashboard
When a human has to weigh options, spot a pattern nobody defined in advance, or defend a decision to a board, a well-built dashboard is still the best tool there is.
Agents are excellent at questions someone already thought to ask. The dashboard is where the questions nobody thought to ask get found.
What we changed, and when
Three things. I share them because other consultancies are having the same conversation, some of them only now.
Step 1: We stopped positioning on the front end alone, in 2025. The Tableau practice has not shrunk, but what it sells widened: semantic models, governed metrics, agent knowledge, MCP endpoints, and the dashboards that sit on top of all of it.
Step 2: We shipped an app before the platform shipped Studio. If we were going to say the first draft should be generated, we should be the ones generating it. And we should do it inside the customer's org, with nothing leaving.
Step 3: We moved the best people forward. The consultants who were our top dashboard builders understand the client's business best. They still build the dashboards that matter, now from a generated starting point. They also do discovery, semantic design and the deployed-engineer work of putting data apps inside real workflows. That is a wider remit, not a consolation.
Frequently asked questions
What are Tableau Data Apps?
Tableau Data Apps are task-specific, interactive applications built on Tableau's governed data and knowledge. They can be generated from a natural language description in Tableau Studio or AI tools such as ChatGPT, then hosted and governed in Tableau, and opened in Slack, Teams, Salesforce, Claude or ChatGPT.
What is the difference between a data app and a dashboard?
A dashboard is a general-purpose view of a subject, built for exploring, comparing and forming a judgement. A data app is built around one task, shows only what that task needs, and usually ends in an action. Both draw on the same governed knowledge.
What is governed vibe coding in Tableau?
Governed vibe coding is Tableau's approach to building apps and dashboards from a plain-language description. Unlike generic AI app builders, the result uses signed-off metrics from Tableau Knowledge, enforces real user permissions, routes actions through the Decision Engine and stays visible in the Command Center.
Will vibe coding replace Tableau developers?
It replaces the hand-built first draft, not the developer. Generated apps are often good up to the last ten percent. Skilled builders still turn a correct draft into one people trust, and their work widens into semantic design, encoded business rules and putting apps inside real workflows.
Where can Tableau Data Apps run?
Through Tableau MCP, data apps can run in Slack, Microsoft Teams, Salesforce, Claude and ChatGPT, with the same governed logic and permissions behind each surface.
Conclusion: the honest ending
I still feel the shift. Fourteen years of a craft do not stop mattering because a keynote said so, and they have not stopped mattering.
What changed is where in the process the craft is applied. I would rather say that in public than sell a client the old process because it is the one we were good at.
Tableau is no longer a BI platform. It is an application platform built on a governed knowledge graph, and dashboards, agents, conversations, alerts and vibe-coded apps all draw from the same graph. On that architecture it currently defines what good looks like.
The dashboard is one of its outputs, still the most important one for a human who has to decide. It is just no longer the only one, and no longer the only thing we are paid for.
If your analytics team is still measured only in dashboards shipped, this is the year to widen the measure. If you want a second opinion on what to add, that is a conversation we have been having for two years, and we are glad to have it with you.
.webp)