Knowledge Hub

Visionaries and business users thought that because they understand their business, know their problems, understand their use cases, and now have Vibe Coding and AI-assisted development at their fingertips, they could build things themselves.

So they began asking:

"Why do we need Data people? We understand the business. We know what we want to build. And AI can help us build it."

And to be fair, the early results look promising.

POCs went well.
Pilots went well.
Applications were built. Dashboards were created. Workflows were automated.

Everyone is excited.

But now comes the interesting part.

• Now it's time to bring in the actual enterprise data.

• Now it's time for proper user acceptance testing.

• Now it's time to fix data quality issues that nobody noticed during the POC.

• Now it's time to optimise the code because compute consumption is increasing, cloud costs are approaching budget limits, and the expected ROI is taking a hit.

• Now it's time to optimise storage, remove unnecessary data duplication, and manage growing volumes.

• Now it's time to integrate with existing enterprise systems, deal with incompatible data structures, and resolve dependencies.

• Now it's time to migrate from legacy technologies to newer platforms without disrupting business operations.

• Now regulatory bodies are asking questions. We need to produce compliance reports, demonstrate data lineage, maintain audit trails, and prove where the data came from and how it was transformed.

• Now business leaders want transparent and trusted KPIs. We need to reconcile different definitions, validate calculations, and ensure everyone is looking at the same numbers.

I have seen organisations invest years building data platforms, introducing new technologies, and expanding their enterprise architecture. Over time, these investments become part of the technology landscape.

Then something interesting happens.

A new platform is introduced, but the old one stays. A new data warehouse is built, but the previous warehouse continues running. A new reporting solution is launched, but business users still depend on the old reports. Another integration layer is added because replacing the existing one is considered too difficult.

Eventually, the organisation is operating multiple generations of technology, often solving similar problems in slightly different ways.

And nobody really wants to switch anything off.

There are understandable reasons. Dependencies are complicated, business users are comfortable with existing systems, migration carries risk, and nobody wants to be responsible for an outage.

Enterprise data architecture is failing. Not because organizations lack investment, or because the problem is inherently too complex, but because the structure required to hold these ecosystems together is either weak or absent. As data landscapes stretch across domains, platforms, and workloads, coherence simply does not sustain itself. Below diagram show the potentiate challenges faced by different persona(s) in an organization.

Over the past decade, there has been no shortage of progress. Cloud-native platforms, lakehouse patterns, distributed processing, governance tooling - none of this is immature anymore. And yet the outcomes tell a different story. Environments expand, but alignment erodes. Governance becomes more visible, more documented, more discussed - but less effective in practice. Analytical capability grows, but trust lags behind, sometimes significantly. These are not isolated execution failures. They point to something more fundamental.