The Conference Beneath the Conference

Every conference has two versions of itself. There’s the version on the keynote stage: the future, the roadmap, the vision, the cinematic trailer for where the industry believes it’s heading.

And then there’s the conference beneath the conference: the hallway conversations, the applause lines, the side comments after sessions, the things practitioners actually lean in for.

The keynote narrative this year was unmistakable. Agentic AI. Automation. Semantic systems. AI-assisted analytics. Composable intelligence. Operational decision-making systems. Tableau was very clearly positioning itself beyond “reporting” and toward becoming more of an operational plane for decision support.

And to Tableau’s credit, the presentations themselves were significantly more coherent this year. Tableau Next now exists as an actual product rather than a conceptual placeholder, while Tableau Cloud, Server, and Desktop were still clearly presented as important products in their own right. There was also a genuine effort to reconnect the company to practitioners and data people, which matters more than many Salesforce executives realize.

But what fascinated me most wasn’t the future everyone was talking about. It was the present moment that everyone was reacting to.

Because when the loudest applause at Devs on Stage comes from:

• dynamic formatting
• vector exports
• layers
• composable data sources
• dark mode
• workflow improvements
• friction reduction

…that tells you something important. It tells you where organizations actually are today.

The Industry Has Entered a Split-Phase Moment

There’s a widening gap between what the vendors are imagining and what organizations are operationally prepared to absorb.

Most companies are still struggling with:

• fragmented business logic
• siloed metrics
• undocumented assumptions
• disconnected systems
• governance confusion
• organizational sprawl

At the same time, the industry conversation has accelerated straight to multi-agent orchestration and autonomous analytics systems.

That’s like trying to install autonomous navigation systems onto a ship whose maps are contradictory and whose instruments disagree with one another.

One of my observations after the keynote was that the companies featured in the demos represented a kind of idealized (perfect world) “all-in” Tableau customer. But most Tableau customers don’t look like that. Most enterprises are messy. They have various BI tools competing internally. Their semantic layers are inconsistent. Their governance is nascent. Their data models have evolved over years of local and siloed decisions with tactical compromises.

Ironically, Tableau’s own strengths have contributed to this reality.

Tableau became wildly successful because it could connect to almost anything, anywhere, regardless of how organized the underlying data actually was. That flexibility enabled enormous creativity and speed. But it also enabled organizations to postpone some of the harder structural work around governance, semantics, and shared definitions.

In other words: Tableau has enabled organizations to avoid doing the difficult work for a very long time. Now AI is arriving, fast. Suddenly all of that deferred structure matters.

AI Is Exposing Structural Debt

One of the most important shifts occurring is that AI is exposing previously invisible organizational problems. For years, companies could operate with undocumented logic, excess headcount, tribal knowledge, discordant communications, dashboard sprawl, and inconsistent semantics. All because human beings are remarkably good at compensating for broken systems.

AI is not. Or more precisely: AI now forces the organization to finally confront the fact that their operating system was already broken.

An AI agent immediately asks uncomfortable questions:

• What exactly does this metric mean
• Which definition is canonical
• Where does this business logic live
• Which source is authoritative
• Can this output be trusted
• Who governs this process

Most organizations discover very quickly that the answers are fuzzier than they assumed. That’s why metadata, ontology, governance, and semantic consistency suddenly matter so much. Not because they’re glamorous. Because they’re load-bearing.

One theme I kept looking for throughout TC26 was the transition from AI excitement to AI governance. We’re still in that excitement and open-questioning phase. The keynote was clearly speaking to a future vision. But when you looked at what practitioners were actually reacting to, you could see that most organizations are still attempting to solve much more foundational problems around workflow, usability, governance, and operational coherence.

And honestly, that’s appropriate. Because governance is more difficult than demos.

The Future Is More Composable Than Monolithic

One of the strongest signals from TC26 was the continued movement toward composability. Composable data sources may end up being one of the most consequential announcements at the conference, even though it didn’t receive the flashy keynote treatment that AI did.

Composability matters because the modern enterprise stack was never one system. It is an interdependent ecosystem:

• warehouses
• APIs
• notebooks
• semantic layers
• BI tools
• AI systems
• governance systems
• operational applications
• custom workflows

No single vendor, or employee, owns the entire environment.

The organizations that will navigate this transition best are not the ones chasing a perfect all-in-one platform. They’re the ones building coherence across heterogeneous environments. That requires interoperability, explicit definitions, flexible infrastructure, information architecture, interoperable governance, and shared protocols.

In other words: the hard stuff.

The industry keeps hoping AI will replace operational maturity. What’s actually happening is the opposite. AI is increasing the value of operational maturity.

One of the more potent questions I asked at the conference centered around GUI-first vs. code-first workflows. Tableau is still heavily oriented around the graphical interface for semantic management and agent configuration. And I believe the broader industry is increasingly moving quickly toward API-first and code-first operational models, where governance, version control, approvals, deployment, and observability all live inside modern software workflows.

That tension is not resolved. And I suspect it’s one of the most important strategic tensions that Tableau now faces.

Tableau’s Real Question

The most interesting question hanging over TC26 wasn’t “What features shipped?” It was: “What is Tableau useful for now?” And I don’t mean that cynically; I mean it earnestly.

Tableau has always been more than a visualization engine. At its best, Tableau reduces the friction between curiosity and understanding. You have a question, you explore, you find the signal, you iterate, and you make rapid sense of something. That still matters. In fact, it likely matters even more now in an AI world.

The strongest reactions from the crowd at Devs on Stage wasn’t toward the autonomous or agentic features. It was toward the new analyst features that reduce friction and improve the human experience of working with data. That’s very telling. Because, while AI can generate infinite dashboards, charts, and summaries quickly, that itself is evidence of the split-phase moment we are now in.

The industry is racing toward AI-generated output, while organizations are still struggling to establish the operational coherence, governance, semantics, and shared understanding required to trust and act on that output.

Which means the scarce resource is no longer production throughput. It’s judgment.

Interpretation.
Context.
Sensemaking.
Decision-making.
Shared understanding.

And this is where I think Tableau still has tremendous opportunity. Not merely as a reporting system, but as a platform that helps organizations to collaboratively reason over an Effective Shared Reality.

What TC26 Actually Revealed

What I walked away from most strongly after TC26 wasn’t hype. It was a growing realization that the analytics industry is entering a maturity transition. One that I honestly would have thought had come before me.

The next decade probably won’t belong to the companies with the flashiest demos. It will belong to the organizations that:

• structure knowledge well
• govern data intelligently
• document workflows clearly
• integrate systems coherently
• reduce friction everywhere
• build systems that humans and machines can reason within together as collaborators

That may sound less exciting than “fully autonomous analytics agents,” but I suspect it’s closer to reality. Because businesses don’t run on vibes. They run on definitions, rules, incentives, constraints, governance, and an effective shared reality.

AI will finally force everyone to make those things explicit.