If 2024 was the year AI captured imagination, 2025 was the year reality caught up.

In this end-of-year conversation, Zhamak Dehghani (Founder & CEO), Cornelius Willis (Head of GTM), and Nextdata engineering leaders Sina Jahan and Jörg Schad reflect on what they actually saw inside large enterprises this year—beyond the hype cycles, demos, and headlines.

What emerges is a clear pattern: AI is moving fast, but data is not. And that gap is now the single biggest constraint on real business value from AI.

👉 Watch the full conversation for the unfiltered discussion.

The Biggest Surprise of 2025: AI Isn’t the Bottleneck—Data Is

Across customers, prospects, and partners, one surprise came up again and again:
it’s not model capability holding companies back.

Enterprises believe AI will work. Budgets are opening. Conviction is high.

But the foundational work, knowing where data lives, who can access it, whether it’s trusted, governed, and usable by machines, is still missing. In some organizations, AI has been banned outright, not because leaders are skeptical, but because the potential blast radius of using untrusted data is too large.

This has created a strange, split reality:

• AI teams racing ahead in isolation
• Data teams stuck maintaining slow, brittle pipelines
• Business teams frustrated by lack of progress

That disconnect is the clearest indictment of Data 2.0 and the modern data stack.

The Most Common Mistake: Skipping the Basics

Despite sophisticated tooling, many enterprises are still missing the fundamentals:

• Data discovery that works in business language, without requiring infrastructure knowledge
• Access controls people can actually understand
• Visibility into duplicated data spend and unused assets
• Clear ownership and accountability at the data product level

In multiple cases, organizations were purchasing the same external data several times across different teams, simply because no one knew it already existed elsewhere.

These aren’t edge cases. They’re structural failures of a model built for storage and pipelines, not for speed, trust, and AI-driven consumption.

From Data 2.0 to Data 3.0: Why the Old Model Is Hitting a Wall

What became undeniable in 2025 is that incremental fixes aren’t enough.

The modern data stack: pipelines, catalogs, warehouses, and access layers wired together, was never designed for:

• multimodal data
• autonomous agents
• real-time decisioning
• or a 50× gap between app development speed and data readiness

As Zhamak puts it in the conversation, this isn’t a “2.1” upgrade. It’s a major version change in how data is produced, governed, and consumed.

That’s the shift we describe as Data 3.0: autonomous data products that collapse pipelines, embed governance by default, and deliver AI-ready data directly to applications and agents.

Governance Is Finally Becoming an Enabler

One encouraging signal from 2025: governance is starting to change shape.

Instead of being a centralized blocker, strong governance, when automated via computation, is increasingly seen as the foundation for speed, the thing that allows teams to move faster because risk is controlled, not despite it.

We’re not fully there yet. But the realization is spreading: without built-in computational governance, AI experimentation simply won’t scale safely.

One Word That Defined the Year

When asked for the word that captured 2025, the answers converged quickly:

• MCP, promising, evolving, imperfect
• Trust, across data, AI, ownership, and outcomes
• Blast radius, what happens when autonomous systems act on the wrong data
• AI-ready data,  widely used, rarely well-defined

That last one matters most. Everyone wants AI-ready data, but few can explain what it actually takes to get there.

What Leaders Should Focus on in 2026

The advice from the team is refreshingly practical:

• Build a stable, governed data foundation that supports rapid iteration
• Measure what matters: speed from data to action, not dashboards or queries
• Tie data work directly to business outcomes: Cost reduction, revenue lift, operational efficiency
• Don’t trade speed for safety; design for both from day one

Noise will only increase in 2026. Clear KPIs are the only way through it.

Why We’re Optimistic

What’s most energizing isn’t the technology, it’s the people.

Across industries, we’re seeing courageous leaders willing to challenge legacy models, break silos, and work differently. The most successful transformations happen when teams go into the trenches together, prove value quickly, and let results, not opinions, carry the change forward

That’s the path we’re leaning into at Nextdata.

Watch the full conversation to hear the nuance, debate, and lived experience behind these insights—and to understand why 2026 may finally be the year data stops holding AI back.

Happy New Year from all of us at Nextdata.