The prevailing boardroom narrative frames data governance as a something birthed by generative AI and the tangled data estates of the 2020s. Having worked through the transitions from legacy mainframe architecture to modern cloud ecosystems, I’d argue this narrative is getting the diagnosis wrong.
Enterprises are not facing a ‘governance gap’ created by new technology. They are experiencing what I call Governance Memory Loss: a slow erosion of discipline that was chosen, incrementally, over two decades. We have treated data governance as a new frontier to be conquered, when in fact it was once the structural foundation of enterprise stability. We didn’t lose the ability to govern, we stopped finding it convenient.
The Mainframe Era: Discipline by Design
In the mainframe era, governance wasn’t a department, a program, or a separate oversight function. It was structural. The architecture itself enforced discipline because deviation was too costly to permit.
Data attributes were centrally defined. There was no ‘shadow data’ because the system simply wouldn’t accommodate it. Diverging from established standards carried a real financial and operational price. Governance was the silent foundation of the architecture, and because it wasn’t a standalone program, it couldn’t be quietly defunded during a budget review.
The ETL Era: Governance as Engineering Output
As enterprises moved into large-scale integration platforms, governance evolved but held its place within the engineering workflow. In the ETL era, discipline emerged naturally from robust technical practice. Data movement was explicit, logic was centralized, and lineage was visible. Transformations were deliberate, often self-documenting to the point where business users could follow the journey of information without needing a translator.
Governance wasn’t a ‘brand’ in those years. It was simply the outcome of building systems with rigor.
The Turning Point
The erosion began not with a technological failure, but with a shift in executive perception. Governance came to be viewed as overhead, a cost center that generated documentation but delivered no immediate commercial return. It was underfunded, deprioritized, and reduced to intent rather than execution. Organizations began articulating what they should do with data, rather than building systems that ensured it happened.
This shift didn’t occur in isolation. As governance was being pushed to the periphery, the regulatory landscape was tightening. The GDPR in Europe, the UK Data Protection Act, and India’s Digital Personal Data Protection (DPDP) Act transformed what had once been a documentation burden into a material strategic risk.
When governance moved out of the engineering workflow and into parallel manual processes, it lost its teeth. It became a post-hoc exercise, a checkbox that could never keep pace with the speed of data creation.
The Cloud Era: Technical Debt by Default
Modern cloud platforms accelerated this abandonment. Elastic compute, low storage costs, and self-service tooling shifted the primary enterprise incentive entirely toward speed. The structural discipline of the mainframe era and the traceable engineering of ETL were framed as friction and discarded accordingly.
Governance became fragmented, manual, and entirely disconnected from the actual data pipeline. The result is technical debt by default: environments optimized for agility at the direct cost of structure, legally and operationally fragile, particularly in markets where multi-platform environments are not just common but a regulatory necessity.
Why the Memory Loss Ends Here
For years, enterprises operated on implicit assumptions, undocumented logic and institutional knowledge held in the heads of long-tenured employees. We could afford to defer the governance problem because human intervention could still bridge the gaps. That buffer is narrowing.
Artificial intelligence cannot operate on implicit assumptions. It requires the context, meaning, and lineage we chose to stop documenting two decades ago. What we are now confronting is the Implicit Assumption Trap: as institutional knowledge erodes, AI performance collapses, not gradually, but at the precise moment it is needed most. Reality has dragged governance back to the centre of the enterprise agenda, not because leaders are asking for it, but because AI is demanding it.
Part 2 examines how AI exposes this trap directly, and why platform migration represents the best remaining window to address it. Stay Tuned!
