Skip to content
Verellax

AI-adjacent work arrives on top of the whole role, never instead of a layer of it.

Skills Index. What the market hires for.

Across 24,944 live specifications, 24,941 carrying at least one mapped skill, as at 14 September 2026, every skill mentioned is assigned to one of four layers (execution, coordination, judgement, AI-adjacent) and one of fourteen sub-categories, and every figure below is the count of specifications carrying that work.

Live basis · W38 · as at 14 September 2026 · 24,944 live specifications · 830 canonical skills
The four layers on the mark, sized by nothing, counted by the data
24,944 live specifications with mapped skills · hover a shape · click to filter the fourteen below
Four layers · 24,944 live specifications Where AI-adjacent work sits 10,199 of the 10,792 live specifications carrying AI-adjacent work carry all three human layers with it, 94.5%

AI-adjacent work arrives on top of the whole human stack, never instead of any part of it. 356 carry it without judgement, roughly one in thirty; none carry it alone.

execution 24,940coordination 24,314judgement 24,141ai-adjacent 10,792
The geometry is the mark's, not the data's: a diagram with counts on it. Area-true, the three human discs would coincide; that finding is stated below in counts.
Skill observations571,156263,750 on live specifications · as at 14 September 2026
Carry AI-adjacent work43.3%10,792 of 24,944 live · 10,199 on top of all three human layers
Carry all three human layers94.5%23,575 of 24,944 · 10,199 add AI-adjacent work, 13,376 do not
Canonical skills830512 execution · 104 coordination · 161 judgement · 53 ai-adjacent
The finding the map has to carry

Of 10,792 live specifications carrying AI-adjacent work, 10,199 carry all three human layers with it, 94.5%. 356 carry it without judgement, roughly one in thirty. None carry it alone.

01

How the four layers are defined

Four layers, fourteen sub-categories. Every classified skill sits in exactly one of each, and every figure on this page is the count of live specifications carrying that work.

Full definitions at /skills/taxonomy
Execution24,940The work a model absorbs first: producing, processing and completing defined tasks.100.0% of live · functional · domain · technical
Coordination24,314The work of moving it between people.97.5% of live · internal orchestration · external orchestration · enablement · strategic alignment
Judgement24,141The work of deciding.96.8% of live · operational judgement · strategic judgement · evaluative judgement · taste
AI-adjacent10,792The work of operating the model itself.43.3% of live · ai tool use · ai tool building · ai governance
02

How many live roles carry each layer

Of 24,944 live specifications as at 14 September 2026, 24,940 carry execution work (100.0%), 24,314 carry coordination work (97.5%), 24,141 carry judgement work (96.8%), 10,792 carry ai-adjacent work (43.3%). The three human layers are present in nearly every role; the fourth is present in 43.3%.

Live specifications carrying each layer
24,944 live specs · a spec carries a layer if one mapped skill falls in it · live basis
Execution100.0% of live
24,940
Coordination97.5% of live
24,314
Judgement96.8% of live
24,141
AI-adjacent43.3% of live
10,792

The three human layers are present in nearly every role; the fourth is present in 43.3%. Area-true, the three human discs would coincide, so the finding is stated here in counts.

Source: Verellax workforce intelligence · 24,944 live specifications, 24,941 with a mapped skill · 263,750 skill observations · 830 canonical skills · live basis · as at 14 September 2026
03

Where AI-adjacent work sits when it appears

Of the 10,792 live specifications carrying AI-adjacent work, 10,199 carry all three human layers with it, 94.5%. 568 carry two human layers, 356 carry it without judgement and 25 with execution alone; none carry it on its own. Across the whole corpus 23,575 specifications carry all three human layers, 94.5%; 10,199 of those add AI-adjacent work and 13,376 do not. The map at the top of this page shows that: hover any shape to see the count.

What sits beneath AI-adjacent work, where it appears
10,792 live specs carrying AI-adjacent work · by the human layers present · live basis
All three human layers94.5% of 10,792
10,199
Two human layers5.3% of 10,792
568
Without judgement3.3% of 10,792
356
Execution alone0.2% of 10,792
25

AI-adjacent work is an addition to a whole role, not a replacement for a layer of one. 356 carry it without judgement, roughly one in thirty. None carry it alone.

Source: Verellax workforce intelligence · 24,944 live specifications, 24,941 with a mapped skill · 263,750 skill observations · 830 canonical skills · live basis · as at 14 September 2026
04

Which sub-categories employers ask for most

Internal orchestration appears in 22,717 live specifications, Functional appears in 20,983 live specifications, Domain appears in 20,577 live specifications, Operational judgement appears in 20,343 live specifications. AI tool use, the largest AI-adjacent sub-category, appears in 8,723; AI governance in 1,596. Employers ask for the use of the model five times more often than they ask for the governance of it.

Fourteen sub-categories, sized by the live specifications that carry them
Four clusters in the four layer colours · hover a node for its definition and count · click to filter the table
execution · 24,940coordination · 24,314judgement · 24,141ai-adjacent · 10,792Functional20,983Domain20,577Technical16,788Internal orchestration22,717External orchestration11,952Enablement9,503Strategic alignment4,806Operational judgement20,343Strategic judgement11,745Evaluative judgement11,008Taste921AI tool use8,723AI tool building2,042AI governance1,596
Hover a node Tool use against governance 5 : 1 Employers ask for people who use the model five times more often than they ask for people who govern it. 8,723 against 1,596 live specifications · the skills-side statement of the decision-architecture finding
Execution24,940 specs
Coordination24,314 specs
Judgement24,141 specs
AI-adjacent10,792 specs
Node area is proportional to specifications carrying the sub-category; positions are fixed by layer and do not encode data. Source: Verellax workforce intelligence · 24,944 live specifications · live basis · as at 14 September 2026
LayerSub-categorySpecificationsShare of live
CoordinationInternal orchestration22,71791.1%
ExecutionFunctional20,98384.1%
ExecutionDomain20,57782.5%
JudgementOperational judgement20,34381.6%
ExecutionTechnical16,78867.3%
CoordinationExternal orchestration11,95247.9%
JudgementStrategic judgement11,74547.1%
JudgementEvaluative judgement11,00844.1%
CoordinationEnablement9,50338.1%
AI-adjacentAI tool use8,72335.0%
CoordinationStrategic alignment4,80619.3%
AI-adjacentAI tool building2,0428.2%
AI-adjacentAI governance1,5966.4%
JudgementTaste9213.7%
Sub-category counts are specifications, not observations: a specification listing three technical skills counts once for technical.
05

How far the mapping reaches, per layer

Mapping completeness is reported per layer, never as one blended number. The AI-adjacent layer is the weakest, and it is the one carrying the thesis; until coverage is flat, skills figures are composition claims, never trends. The bars are carried from W36 and print that read date until the definition is pinned.

Skills mapping coverage, per layer
canonical specifications · never blended · carried from W36 · as at 4 September 2026
Execution
85.3%
Coordination
71.6%
Judgement
70.2%
AI-adjacent
67.7%

The AI-adjacent layer is the weakest and it is the one carrying the thesis. Until coverage is flat, skills figures are composition claims, never trends.

Source: Verellax workforce intelligence · 830 canonical skills · 571,156 skill rows · carried from W36 · as at 4 September 2026
06

How the skill mix differs by market

Execution load runs from 42.9% of the work in Singapore to 49.7% in Spain. The market pages carry each market's split.

The markets
Execution load against judgement load, by market
ten markets · average share of each role's work · live basis
Spain
49.7%
20.1%
Australia
48.1%
22.0%
India
48.1%
20.2%
Ireland
48.0%
20.7%
United States
47.8%
21.6%
Canada
47.2%
20.6%
France
47.1%
20.0%
Germany
45.2%
21.7%
United Kingdom
44.5%
23.0%
Singapore
42.9%
23.9%

Execution load runs from 42.9% of the work in Singapore to 49.7% in Spain.

Execution, the layer a model absorbs firstJudgement
Source: Verellax workforce intelligence · live specifications, ten markets with a layer read · as at 14 September 2026
07

How this is measured

Live basis. A specification carries a layer if at least one of its mapped skills falls in that layer. Sub-category counts are specifications, not observations; a specification listing three technical skills counts once for technical. Canonical skills are those with taxonomy status active.

Full method at /method
Source: Verellax workforce intelligence · 24,944 live specifications, 24,941 with a mapped skill · 263,750 skill observations · 830 canonical skills · live basis · as at 14 September 2026
Figures re-render each Monday from that week's pack.