AI is commonly framed as a common equaliser.
Give everybody entry to the identical instruments. Present widespread coaching. Increase the baseline—and outcomes ought to converge.
However this assumption is more and more misaligned with actuality.
Regardless of unprecedented entry to AI instruments and studying sources, outcomes are diverging—not converging. A small group compounds quickly, a center group makes incremental good points, and an extended tail struggles to transform entry into significant progress.
This isn’t only a productiveness concern. It’s a query of structural divergence at scale—with implications for financial mobility, workforce resilience, and social stability.
To know why, we have to transfer past entry and study what actually governs outcomes.
The hidden variable: Structural friction
The core constraint shouldn’t be entry. It’s structural friction.
Structural friction is the overall resistance a person faces in changing entry into functionality—throughout cognitive, environmental, and behavioural dimensions.
This consists of:
The psychological fashions people convey
The environments they function inside
The habits, incentives, and identities they carry
Their present capabilities
Two people could also be equally uncovered to AI. They might have entry to the identical instruments, programs, and knowledge.
But their skill to transform that entry into significant outcomes can differ dramatically.
Why?
As a result of the hassle required to maneuver ahead shouldn’t be fixed. It varies—typically subtly, typically exponentially.
Additionally Learn: What I inform my children to have the ability to thrive within the age of AI
From distance to gradient
Most discussions of AI functionality give attention to distance:
How far somebody is from the technological frontier
How a lot they should be taught
However distance is the improper lens.
What issues is the gradient—the hassle required to shut that distance.
Curve A: The steep precipice (Mature, non-digital SME)
For these formed by many years of deterministic workflows, even the primary shift requires a cognitive leap. There is no such thing as a gradual incline—solely discontinuity.
Curve B: The linear slope (Mid-career PMET)
For professionals in structured, digitised environments, the climb is regular. Company coaching scaffolds the journey. Progress is actual—however bounded by context.
Curve C: The flat runway (AI-Native):
For these with computational or design-native backgrounds, transferring throughout AI layers feels intuitive. The friction is minimal. Every new software maps naturally onto current psychological fashions.
Two people could also be equally removed from the frontier. One faces a manageable slope. The opposite faces a near-vertical wall.
The distinction shouldn’t be seen in entry. It’s embedded within the construction.
When outcomes comply with an influence legislation
As soon as we account for the gradient, a well-known sample emerges: power-law outcomes.
In such techniques:
A small group captures disproportionate worth
A center group sees incremental enchancment
A protracted tail struggles to translate effort into outcomes
That is the place most AI upskilling narratives break down.
They assume that uniform inputs—programs, subsidies, software entry—will produce broad-based uplift.
However power-law techniques don’t reply to uniform inputs. They amplify preliminary variations.
The identical intervention that accelerates the highest 10 per cent might depart the underside 50 per cent largely unchanged.
Additionally Learn: Chew-sized innovation: A sensible path for SMEs to maintain development
The compounding impact of velocity
Even this framing is incomplete.
If structural friction have been the one constraint, time and persistence would possibly nonetheless shut the hole.
However AI doesn’t stand nonetheless.
What’s present at this time turns into out of date in months. Interfaces evolve, paradigms shift, and expectations reset constantly.
For these on low-friction curves, velocity compounds benefit. They adapt in actual time and prolong their lead.
For these on high-friction curves, velocity is destabilising. Simply as a foothold is gained, the bottom shifts once more.
The consequence is not only a widening hole—it’s a transferring goal with accelerating divergence.
From hole to transferring staircase
When adaptation constraints are non-linear, the dynamics change essentially.
These dealing with decrease friction transfer quicker
The technological frontier advances concurrently
The space to relevance doesn’t shrink—it shifts
This creates what will be described as a transferring staircase impact: Some are climbing. Some are standing nonetheless. However the staircase itself is rising.
The hole doesn’t merely widen. It accelerates.
Upskilling is not about catching up. It’s about chasing a horizon that retains receding—and questioning whether or not catching up nonetheless issues by the point one will get there.
Over time, if this dynamic persists, we might even see the emergence of disengagement behaviours—sub-cultures of “mendacity flat(躺平)” to withdrawal(ヒキコモリ)—bringing broader social and coverage implications.
Why most AI efforts converge to common
With no clear understanding of those dynamics, people and organisations reply predictably:
They attempt to transfer quicker. They undertake instruments aggressively. They experiment constantly.
However with out correct analysis, this typically ends in:
Poor downside definition → quicker improper options
Shallow understanding → extra polished however common outputs
Misaligned objectives → environment friendly irrelevance
AI doesn’t repair flawed pondering. It scales it.
The failure of one-size-fits-all options
Most interventions stay structurally uniform:
Standardised programs
Broad-based subsidies
Generic “AI literacy” programmes
These approaches assume that entry is the first constraint.
It’s not.
The actual constraint is misdiagnosis.
In a power-law system, the common answer serves nobody:
Too shallow for individuals who want transformation
Too gradual for individuals who are already accelerating
Misaligned for these dealing with structural discontinuity
Additionally Learn: Employees dash forward of bosses in AI adoption in Singapore, exposing a metamorphosis hole
Reframing the issue: From content material to cognition
If the problem is structural, the answer can’t be purely curricular.
We should shift from what we educate to how people suppose and function.
This requires a extra exact diagnostic lens:
What psychological fashions does the learner already possess?
How adaptable is their working setting?
How a lot unlearning/relearning is required earlier than new capabilities can take maintain?
For some, the barrier is technical data. For others, it’s id, behavior, or worldview. For a lot of, it’s the interplay of all three elements.
Till these elements are surfaced, upskilling efforts will stay blunt devices.
Designing for divergence, not uniformity
A simpler strategy begins with divergence because the baseline.
It should account not just for functionality, but in addition for intent.
An skill–motivation matrix offers a sensible strategy to operationalise this:
Group
Description
What they want
The natives
Excessive skill, excessive motivation
Take away constraints, maximise leverage, allow fast experimentation
The integrators
Average skill, excessive motivation
Redesign workflows, embed AI into current techniques, and scale back integration friction
The constrained adopters
Low skill, excessive motivation
Simplify instruments, scale back cognitive load, present guided pathways
The at-risk
Low skill, low motivation
Reframe relevance, rebuild motivation, handle id and behavioural limitations
The sooner gradient mannequin explains why outcomes diverge. This matrix offers a strategy to act on that divergence.
Implications: Past upskilling
For policymakers
Upskilling stays obligatory—however it’s inadequate by itself.
Equal entry to coaching doesn’t produce equal outcomes when structural friction varies. Interventions should account for beginning circumstances, not simply curriculum supply.
For founders and builders
The following wave of worth lies not in constructing extra AI instruments—however in lowering friction.
Alternatives lie in:
Abstraction
Simplification
Humanising interfaces
Not everybody can transfer quicker. However techniques will be designed to require much less motion.
For economies
In extremely related and urbanised societies, uneven acceleration introduces structural danger.
It erodes the steadiness of the standard bell curve and creates sharper divides.
On the similar time, sustaining competitiveness requires steady enchancment of the nationwide common.
How these gradients are managed will form each financial efficiency and social cohesion.
Additionally Learn: Washington switched off the world’s strongest AI, and Asia must be paying consideration
Singapore as a baseline
Singapore’s coordinated and well-funded strategy offers a helpful baseline.
It doesn’t invalidate the gradient framework—it reinforces it.
If structural friction emerges even in a high-trust, policy-aligned setting, it means that comparable dynamics will manifest extra strongly elsewhere—significantly in areas with fewer institutional buffers and steeper cognitive limitations.
The underside line
AI has the potential to be both a common equaliser or a robust divider.
Entry is critical—however removed from enough.
In a power-law system, equality of enter doesn’t produce equality of end result. Structural friction determines who compounds—and who stalls.
We are able to perceive this by a Energy-Legislation Constraint Mannequin:
Outcomes are formed by the interplay between structural friction, compounding dynamics, and a constantly shifting technological frontier.
The divide is not going to be closed by giving everybody the identical curriculum.
The danger shouldn’t be that AI leaves folks behind. It’s that it accelerates inconsistently, and we mistake that acceleration for progress.
Till we be taught to see the gradients, we are going to proceed to optimise for movement somewhat than significant enchancment.
Over time, these gradients don’t disappear. They accumulate—settling into the grains of structural inequality.
In a power-law system, progress is set not by how far you’re, however by how onerous it’s so that you can transfer, and how briskly the world is transferring when you attempt.
—
Editor’s word: e27 goals to foster thought management by publishing views from the group. You can too share your perspective by submitting an article, video, podcast, or infographic.
The views expressed on this article are these of the writer and don’t essentially mirror the official coverage or place of e27.
Be part of us on WhatsApp, Instagram, Fb, X, and LinkedIn to remain related.
The submit The grains of structural friction: Why AI upskilling is failing the ability legislation appeared first on e27.
