Old version – Spaces and Guardrails

Modern organizations increasingly promote two mantras:

• “Everyone must upskill in technology / data / AI.”
• “Data scientists must become business partners.”

Both statements are not only naïve — they are cognitively incoherent.
They confuse collaboration with convergence, and integration with assimilation.

What is at stake is not skill acquisition or mindset, but the integrity of distinct cognitive regimes.

1. The False Promise of Cognitive Convergence

The idea that business experts and technology / AI experts can — or should — converge toward a single mode of thinking rests on a flawed assumption:

that intelligence is homogeneous, transferable, and context-free.

Decades of research say the opposite.

Tacit vs. Abstract Knowledge

Michael Polanyi’s work in The Tacit Dimension established a fundamental distinction:

“We know more than we can tell.”

Business expertise — particularly in finance, operations, regulation, risk, or supply chains — is:

• situated in concrete contexts,
• historically accumulated,
• deeply constrained by exceptions, edge cases, and institutional memory,
• only partially formalizable without loss of meaning.

This knowledge does not scale by abstraction.
It scales by exposure, repetition, and judgment under constraint.

Technology and AI expertise, by contrast, is:

• abstract by design,
• composable and modular,
• detached from local meaning,
• optimized for recombination and transfer across domains.

These are not two ends of a learning curve.
They are orthogonal epistemologies.

2. Two Rationalities, Two Truth Regimes

Herbert Simon’s concept of bounded rationality in Rationality in Organizations, later extended by Gerd Gigerenzer in The Adaptive Toolbox, makes the distinction sharper.

Business Cognition: Ecological Rationality

Business experts operate under:

• irreversible consequences,
• institutional accountability,
• legal, financial, and reputational constraints.

Their cognition is:

• heuristic, not algorithmic,
• optimized for robustness, not elegance,
• intolerant to “mostly correct” solutions.

Truth is contextual, situated, and often conservative — for good reason.

Technology & AI Cognition: Analytical Rationality

Technology and data experts operate under:

• reversible experimentation,
• fast feedback loops,
• low marginal cost of error.

Their cognition is:

• model-driven,
• probabilistic,
• exploratory by nature.

Truth is provisional, statistical, and iterative.

When organizations ask business experts to “think like data scientists”, they erode robustness.
When they ask data scientists to “think like the business”, they kill exploration.

3. The Myth of the “Business-Partner Data Scientist”

The injunction “data scientists must become business partners” is particularly damaging.

It implicitly assumes:

• that business understanding is mainly semantic,
• that domain expertise can be absorbed through exposure,
• that abstraction loss is acceptable.

All three assumptions are false.

A data scientist who internalizes business constraints too deeply:

• narrows the hypothesis space prematurely,
• overfits models to legacy logic,
• reproduces existing blind spots at scale.

Conversely, a business expert forced into analytical abstraction:

• loses sensitivity to edge cases,
• underestimates second-order effects,
• mistakes statistical confidence for institutional validity.

The result is not hybrid intelligence, but cognitive dilution.

4. Cognition Is Distributed — Not Blended

Edwin Hutchins demonstrated in Cognition in the Wild that effective systems do not rely on individual intelligence, but on distributed cognitive architectures.

Performance emerges from:

• specialization,
• interface quality,
• clear boundaries between cognitive roles.

At the age of AI, cognition becomes even more composite:

• humans + machines,
• abstraction + judgment,
• speed + constraint.

The problem is not diversity of cognition.
The problem is unclear cognitive sovereignty.

5. Cognitive Spaces: Legitimate Domains of Cognition

A cognitive space is not a role, a function, or a skillset.
It is a domain of legitimate authority over meaning, truth, and decision.

Examples:

• Business experts own sense-making under constraint.
• Technology experts own exploration of the solution space.
• AI systems own statistical regularities at scale.

Each space operates under:

• different validation criteria,
• different time horizons,
• different error tolerance.

None can replace the others.
None should be absorbed by the others.

6. Guardrails: Preventing Cognitive Colonization

Transformation fails when organizations allow one cognitive regime to dominate outside its domain.

Typical failure modes:

• industrializing business perception before it is stabilized,
• forcing data scientists to validate models against intuition rather than evidence,
• demanding that business owners “sign off” on abstractions they cannot cognitively assess.

Cognitive guardrails exist to prevent these errors.

Examples:

• Do not automate perception that has not been collectively clarified.
• Do not require domain intuition from abstraction specialists.
• Do not collapse exploratory and accountable decision spaces.

Guardrails are not bureaucracy.
They are structural protection of collective intelligence.

7. A Non-Negotiable Conclusion

Business and technology experts do not differ by seniority, mindset, or willingness to change.
They differ by cognitive regime.

At the age of AI, organizational performance does not depend on cognitive alignment,
but on the deliberate composition — and protection — of heterogeneous cognitive spaces.

Transformation does not require everyone to think alike.
It requires knowing who must not.