Dashboards were never the destination

For years, dashboards became the proof that a company was “data-driven.” If you built beautiful visualizations with filters and drill-downs, you were seen as modern. But dashboards were never the end state. They were a workaround created because data was messy, tooling was restrictive, and asking open-ended questions of the warehouse wasn’t possible without coding.

They looked like self-service, but they rarely produced decisions.

I’ve built more dashboards than I can count, and I can name very few strategic decisions that hinged on one. What they did best was report what happened and generate more questions. Now, don’t get me wrong, I loved building them, being creative, and solving complex data problems. However… I learned that dashboards were a starting point for end users — inbound ad hoc queries were soon to follow.

Parameters, filters, and drill paths were our attempts to mimic human curiosity, but curiosity doesn’t stop at drop-downs. It follows threads, asks “why,” hops across sources, and changes direction based on what it discovers. Dashboards were never built for that. They surfaced signals, not reasoning. The other stuff — root-cause analysis, causal reasoning, or investigative depth — lived elsewhere in SQL, in notebooks, in analysts’ heads.

And because the reasoning never lived inside the systems we gave the business, it stayed slow, manual, and dependent on humans translating logic on demand.

In short, dashboards create more questions than they answer.

We built towards surfaces instead of systems

By optimizing for dashboards, organizations inherited predictable problems:

• weeks-long turnaround times for analytics requests

• shallow insights that leave a business user stuck

• cycles spent cleaning data outside of the warehouse

• overly complex experiences engineered for all possible questions

We built delivery surfaces, not analytical systems. That’s why dashboard “self-service” never fully worked. Dashboards show fragments. They don't follow the investigative path a person takes when they truly want to understand something.

The market is now acknowledging what most data leaders already know: dashboard-based self-service plateaued. It provided visibility, not analysis. Agentic analytics is the structural evolution of that gap — reasoning systems governed through context, operating directly on the warehouse, and trained by the analysts who need to scale.

A new role for data teams

In a dashboard-first world, analysts spent most of their time building artifacts. In an agentic world, their leverage shifts. They become stewards of context — teaching systems where truth lives, how logic works, and what the business actually means when it asks a question.

The embedded analyst can train an agent to support a function instead of living in a queue. Deep work that once sat in short-staffed tiger teams can scale. And business users can finally move beyond static reporting into genuine answers.

Inside Hex, we’re already seeing this: fewer dashboards, faster investigations, and meaningful reductions in duplicated effort. When the context is correct, the system does the heavy lift.

Respect dashboards, but stop treating them as the goal

Dashboards still matter. They’re excellent for reporting KPIs, surfacing operational signals, and aligning leaders around shared metrics. They allow data teams to be creative in how they display these metrics and will remain a primary surface for need-to-know numbers.

But they’re not analytical engines. They don’t reason, explore, or explain why something happened. They were the best tool available for twenty years — and they served that role well.

We trained business users to ask for dashboards because it was the only language the industry gave them. Now we have better tools and better expectations. Dashboards should support visibility. Context-driven agents should deliver insight.

Dashboards were never the destination.

Reliable, timely, context-aware answers were.

And maybe now, we can finally deliver them.