The Great Reset: How AI Is Forcing Data Teams Back to Zero

In 2017, Joe Reis predicted a shift beyond data-driven organizations to intelligence-driven ones—systems where machines don’t just inform decisions but take actions automatically. At the time, he described the future as “pretty exciting stuff.”

Eight years later, his tone is different.

“Everyone’s sort of collectively staring into the void… we’re all back at the same starting line again.”

In this episode of Data Talks On The Rocks, Joe joins Rill Data Co-Founder & CEO Michael Driscoll to break down what he calls The Great Reset—a moment where AI shockwaves, consolidation, and collapsing toolchains have erased the advantages data teams spent a decade building.

AI Shockwaves and the Return to Zero

Across the industry, AI isn’t just changing workflows—it’s forcing teams to renegotiate the fundamentals of how they build. In data engineering especially, Joe argues the impact has been unusually abrupt:

• Practices once considered standard no longer make sense
  
  
• Toolchains have shrunk instead of expanded
  
  
• Hiring signals have inverted
  
  
• Teams aren’t “stack-first” anymore—they’re outcome-first
  

It’s not an iteration. It’s a reset.

The assumptions that guided the Modern Data Stack era—composability, orchestration, and tool specialization—are being rewritten in real time. And whether teams admit it or not, everyone is relearning the basics under new constraints.

Consolidation and the End of the Modern Data Stack Era

Midway through the conversation, Michael asks a question many leaders are quietly thinking:

“Did the Modern Data Stack ever actually work?”

Joe’s answer is nuanced. The MDS era solved real problems—especially in centralizing analytics in cloud warehouses. But it also created a fragmented ecosystem that scaled operational complexity faster than value.

Today, the pendulum has swung sharply in the opposite direction:

• Platforms are consolidating
  
  
• Interfaces are unifying
  
  
• Teams want fewer moving parts, not more
  
  
• Strong defaults beat infinite choice

Joe captures the shift bluntly:

“At this point, you’ve gotta pick a side — Databricks or Snowflake.”

It’s not tribalism. It’s pragmatism.

Organizations have learned that maintaining a sprawling, hyper-modular toolchain is a cost few can justify—especially when AI is reshaping the upper layers of analytics anyway.

Dark Matter Data: The Unseen Layer AI Still Depends On

One of Joe’s most important observations is about the invisible substrate beneath every “official” stack:

“Dark matter data… Excel sheets, shadow pipelines… that’s what makes the world go around.”

Most organizations architect for their warehouses, lakes, and dashboards.
But the real work often happens in untracked:

• spreadsheets
  
  
• logic buried in ad-hoc scripts
  
  
• one-off transformations
  
  
• business rules no one formally models
  

This hidden layer is where decisions get made—and where AI systems will inherit their strongest biases and biggest risks.

Ignoring it isn’t an option anymore.

Why Craftsmanship Matters More in the AI Era

AI can now generate SQL, documentation, transformations—even architecture scaffolding. But this abundance doesn’t remove the need for human judgment. It makes it more important.

As Joe puts it:

“The marginal cost of another line of code is zero… so craftsmanship is going to matter more, not less.”

When any system can produce infinite code:

• The structure matters more than the syntax
  
  
• The semantics matter more than the pipelines
  
  
• The developer experience matters more than the tooling diagram
  

Craftsmanship today looks like:

• Choosing sensible defaults
  
  
• Designing abstractions humans can understand
  
  
• Defining clear, durable concepts (“customer,” “order,” “event”)
  
  
• Creating systems someone else—or an agent—can safely navigate
  

AI accelerates output. But it can’t decide what “good” looks like. Taste, clarity, and intentional modeling become differentiators.

What Agents Could Mean for Analytics

In the final part of the conversation, Michael and Joe explore the rise of agentic systems—tools that can query, summarize, model, or even initiate workflows autonomously.

These agents introduce new design questions:

• How do semantics evolve when natural language becomes the interface?
  
  
• How do teams govern AI-written code and transformations?
  
  
• What shifts when analysts collaborate with agents instead of operating alone?
  

Joe’s view is clear: agents won’t replace data teams. But they will change what those teams spend their time doing. Execution becomes cheaper. Intent becomes the new constraint.

The Reset is an Opening

Despite the uncertainty – shrinking job markets, collapsing toolchains, AI-driven anxiety – Joe ends on a surprisingly optimistic note:

“Instead of waiting to see what’s on the other side of this void – go build it.”

Every company is reassessing its infrastructure.
Every team is renegotiating its workflows.
Every engineer is reconsidering which skills matter.

This alignment almost never happens.
Teams that embrace the Reset—not resist it—will help shape the next decade of data practice.

Watch the full conversation with Joe Reis and Michael Driscoll here.