Data's Second Act: From Big Dreams to Messy Reality
- By Winston Thomas
- January 18, 2025

In 2014, Mahesh Kumar penned a Harvard Business Review article predicting that big data needed to grow up. A little more than a decade later, it’s clear that adolescence has been messier than anyone expected.
The numbers were staggering back then: USD44 billion projected spending on big data, with USD37.4 billion going to IT services. Today’s figures would make those look quaint, but the fundamental challenge Kumar identified remains: we’re still figuring out how to industrialize our data operations.
“The situation has gotten so complex,” reflects Kumar, the chief marketing officer of Acceldata. “Tens of thousands of databases, literally in the enterprise, many thousands of pipelines, hundreds of thousands of tables and millions of columns.” The dream of data maturity has collided with an exponentially more complicated reality.
The quality crisis nobody talks about
Here’s the uncomfortable truth keeping chief data officers anxious: Major enterprises are self-reporting data quality levels in the “high single digits to teens,” Kumar reveals. Putting it into context, it means the world’s largest companies, despite billions in investment, are achieving reliable data quality barely above 10%.
The advent of GenAI has made this chronic condition an acute one. As Kumar notes, “AI is making it worse from a business outcome perspective because enterprises are leaning more into these types of models for actual, real-time near real-time decisions.”
This leads to a paradox in which AI exacerbates and solves the data quality problem. Kumar describes a large bank that discovered their credit score pipeline had been broken for two months — a critical failure in their lending decisions. But AI also helped the bank understand the data quality issue and detect anomalies.
“AI is both challenging and at the same time is helping solve that problem," Kumar explains. The tools that make data environments more complex are also becoming sophisticated enough to monitor themselves.
Unfortunately, the paradox is creating what Kumar calls a “propensity to just leave it as it is” — companies paralyzed by the complexity of their data environments, responding to crises rather than preventing them.
The data trust revolution
Perhaps the most surprising evolution isn’t technological — it’s organizational. Leading data providers are fundamentally reimagining trust through what Kumar calls “quality scores.” One global data provider, sourcing from 130 countries, now attaches confidence metrics to its data products. It’s even turning the lens back on its suppliers, scoring the quality of incoming data and calculating the true cost of data cleanup.
This shift represents a mature acknowledgment that perfect data is impossible, but quantified reliability is achievable. It is also creating new waves in our approach to data governance.
“The governance approach is due for a complete shakeup,” Kumar argues. The old model of committees and weeks-long approval processes is collapsing under the speed of AI innovation.
“You’re gonna have governance almost like portable governance that moves with the data," Kumar predicts. Think of it as governance-as-code, embedded directly into data pipelines rather than enforced from above.
Decoding data’s next decade
Ten years after Kumar’s original call for data to grow up, we’re seeing signs of maturity, but not in the ways anyone expected. The industrialization of data isn’t following the neat manufacturing analogies of the past. Instead, it’s evolving into something more organic: self-monitoring, self-governing systems that augment human capabilities rather than replacing them.
For CDOs and CMOs, the message is clear: the future of data management isn’t about perfect control but building resilient systems that can intelligently handle imperfection. The companies that thrive will embrace this reality, investing in automation while maintaining human oversight where it matters most.
As Kumar puts it, “That whole process is being reinvented right now in the enterprises, and we’ll see more of that in the next 2, 3 to 5 years.” The data revolution isn’t over — it’s just getting started.
Image credit: iStockphoto/gemenacom
Winston Thomas
Winston Thomas is the editor-in-chief of CDOTrends. He likes to piece together the weird and wondering tech puzzle for readers and identify groundbreaking business models led by tech while waiting for the singularity.