The epistemic debt nobody tracks

The epistemic debt nobody tracks

Technical debt is visible.
Epistemic debt is not.
And yet epistemic debt determines where organisations break long before technology does.

Legacy IT systems are the largest reservoirs of epistemic debt.
Not because they are old, but because no one remembers why they were built, what assumptions they encode, or which decisions they silently enforce.
They run, so they are allowed to keep running.
But every day they run, they accumulate forgotten rationale — a slow collapse of decision memory.

Over time, these systems stop being technology and become institutional mythology.
People inherit them the way you inherit an old building: you don’t know why the walls are where they are, but you learn to walk around them.
The original intent disappears, replaced by workarounds, interpretations, and habits.
The system still functions, but the organisation no longer understands the logic it is built on.

Human actors add another layer.
Roles change, teams rotate, and definitions drift as people reinterpret them to fit their immediate context.
What was once shared meaning becomes a collection of private translations.
Institutional knowledge dissolves into personal memory, and personal memory dissolves into silence.
The organisation keeps moving, but the epistemic continuity that once held it together is gone.

Faulty compliance makes it worse: organisations become compliant with the wrong things because they no longer remember the rationale behind the right ones.

AI does not politely ignore missing rationale.
It does not smooth over semantic drift.
It does not reconstruct lineage that was never documented.
AI actors reveal the gaps as soon as they engage.
Where humans compensate, AI collapses.
It breaks exactly at the points where the organisation has been forgetting the longest.

This is why AI adoption feels disruptive: it forces organisations to confront the epistemic debt they have normalised.
The debt was always there — buried in legacy systems, undocumented decisions, shifting definitions — but AI makes it operationally visible.
Suddenly, what was tolerable becomes a constraint.
What was invisible becomes expensive.

Epistemic debt accumulates quietly.
In definitions that no longer match reality.
In processes built to compensate for missing memory.
In dashboards that present fragments without context.
In decisions whose rationale evaporated years ago.
It is the debt of meaning, not code.

The real modernisation challenge is not cloud migration or platform consolidation.
It is the reconstruction of epistemic continuity in systems that have been drifting for decades in some cases.
Until that happens, AI will continue to reveal what organisations have forgotten, not what they know.

Epistemic debt is not a technical problem.
It is an architectural problem.
And architecture decides whether an organisation can scale, or merely continue to operate.