Cubit in Action

Most AI impact studies assign a single number to a job title. “Financial Analysts face 47% automation risk.” That headline sounds precise, but it is practically useless — and often actively misleading.

A Financial Analyst performs dozens of distinct activities. Some are highly procedural and digital: gathering data from financial systems, building valuation models, running scenario analyses. Others require nuanced judgment: interpreting ambiguous market signals, presenting findings to stakeholders, negotiating with counterparties, building client relationships. These activities have radically different exposure profiles.

A job-level score obscures the fact that the role is already being restructured: some tasks automated, others augmented, others elevated in importance precisely because the routine work has been offloaded. A single number cannot distinguish “automate the charting” from “protect the client relationship.” Under these circumstances, a bird's-eye view leads to a crash, not clarity.

Task-level analysis enables precision that job-level analysis cannot.

It reveals which specific activities within a role face pressure, why, and what factors require human involvement. This transforms abstract “AI risk” scores into actionable workforce strategy. When decision-makers need aggregate job-level or regional metrics, Cubit provides them — but as thoughtful rollups of task-level intelligence, not crude approximations that treat heterogeneous work as monolithic.

The demonstrations that follow are organized around this principle. Each begins with a question existing data cannot answer with sufficient precision, then shows how Cubit's task-level decomposition produces a concrete, actionable answer.

The challenge: A state workforce agency must allocate transition funding across regions. It needs to know which occupations in which metros face pressure, how much employment and wages are exposed, and what transition pathways exist. No existing public dataset connects AI exposure scores to regional employment data; Cubit does.

Regional Scan

Querying California MSAs by at-risk wages reveals substantial variation. At-Risk Jobs estimates positions exposed to automation pressure; At-Risk Wages captures the dollar value; High-Risk Jobs restricts to workers in the most severely exposed occupations.

San Jose leads not because it has the most workers, but because its high-wage technology workforce creates outsized exposure: $1.4 billion in at-risk wages across fewer than 8,000 positions. San Francisco concentrates 1,080 high-risk workers — nearly three times San Diego's 390 despite fewer total employees. That 2.8× difference would be invisible in any aggregate statistics.

MSA Deep Dive

Drilling into San Jose identifies the most exposed occupations. Auto Score (0–100) measures overall susceptibility:

Database Architects score highest (58.3), with 930 workers and $24.9M in at-risk wages. Power Plant Operators score nearly as high (56.8) but employ only 40 workers, while Computer Occupations (15,550 workers, $177M) dominate aggregate exposure despite a lower score — illustrating why both per-occupation and per-region granularity matter.

Task Decomposition

The question is not whether Database Architects are “at risk,” but which tasks drive the exposure. Each task carries four dimension scores producing AI Exposure Potential and Human Imperative. Of 20 scored tasks, 15 (75%) fall in the Automation zone.

These tasks score 8–9 on Procedural Intensity and Digital Accessibility, 1–2 on Physical Embodiment: structured, fully digital work. Zero tasks fall in the Augmentation or Human-Centric zones.

Transition Pathways

Cubit's transition engine matches the source occupation's 120-dimensional requirement profile against the full taxonomy, filtering for positive resilience gain:

Regulators cannot act on “58 out of 100.” But 930 Database Architects earning $24.9M in at-risk wages, with 75% of tasks in the Automation zone and 94% skill overlap with Software Developers — that is the basis for a funded transition program.

The challenge: A mid-size financial services firm (~5,000 employees, ~40 roles) needs to know which roles to automate, which to augment, and how its industry-specific priorities differ from generic assessment. No other dataset supports reweighting; Cubit does.

Portfolio Scan

A batch lookup of 15 representative roles produces the baseline landscape. Resilience (0–100) measures resistance to AI displacement; BIS (Balanced Impact Score) is the net difference:

Custom Scoring

The firm expresses its strategic priorities through dimension weights: amplified Socio-Emotional Depth (client trust is paramount), suppressed Physical Embodiment (irrelevant for office roles). The results invert dramatically:

Personal Financial Advisors jump from 14th to 1st. A high custom score signals strategic relevance: the role's socio-emotional demands become the firm's most important asset to protect. Data Entry Keyers drop from 2nd to last — their procedural intensity no longer differentiates.

The Financial Analyst: Task-Level Reality

Financial Analysts rank mid-table on both standard and custom scoring, yet the paper's opening used them as a conceptual example of internal heterogeneity. Here, the data makes the argument empirical. Of 20 scored tasks: 4 fall in Automation, 14 in Status Quo, and 2 in Human-Centric — contradicting any blanket “at risk” characterization:

Charting and valuation: Procedural 8–9, Digital 9, Socio-Emotional 2. These are automation candidates. “Assessing companies” requires Physical Embodiment of 7 (site visits). “Creating presentations” scores 8 on Socio-Emotional Depth. A firm automates the charting, monitors Status Quo tasks, and doubles down on client-facing work.

Role Comparison: Who Can Transition?

The firm now knows which tasks to automate and which to protect. The next question: what to do with the people in roles being restructured? Keystone skill overlap reveals who can cross-train:

Financial Advisors share 42.9% of keystone skills with Analysts — enough common ground for cross-training. Data Entry Keyers share only 11.8%. Their scores look superficially similar (Auto within 7 points, both with positive BIS), but the skills are almost entirely disjoint. Combined with declining employment, the data tells the firm: Data Entry Keyers are automation candidates, not retraining candidates.

Under standard scoring, the firm would deprioritize its most strategically important client-facing roles — the very positions where human presence creates irreplaceable value — while investing automation resources in procedural work that barely differentiates the business. The same data, scored through a different strategic lens, produces a fundamentally different priority ordering.

The challenge: A workforce board serving a manufacturing community learns a major employer is restructuring. It has funding for one retraining program and needs a target: an occupation displaced machinists can reach, that is growing, and that resists further AI disruption. No existing dataset answers: where should displaced workers go, and what should they learn?

Assessing Occupational Risk

Machinists score 50.2 on automation susceptibility with slower-than-average growth. The number alone is ambiguous — roughly average. Task decomposition reveals why. Of 20 scored tasks, 9 (45%) fall in the Automation zone, 6 in Augmentation, 4 in Status Quo, and 1 in Human-Centric:

All machining tasks are highly procedural (17 of 20 score 8+), so Procedural Intensity alone does not distinguish risk. The differentiator is Physical Embodiment: CNC programming scores 9 on Digital Accessibility and 4 on Physical Embodiment (replicable by software), while hands-on alignment scores 9 on Physical Embodiment (durable). The occupation's physical tasks are protected; its digital tasks are not.

Identifying Adjacent Roles

The transition engine compares skill profiles against every other occupation, filtering for targets where workers gain resilience:

All five candidates share 88–92% of their requirement profile with Machinists — these are not career changes, they are lateral moves. Wind Turbine Service Technicians stands out: highest resilience gain (+6.9), “much faster than average” growth, and a clean energy sector that strengthens the case for public funding.

Mapping the Reskilling Path

Wind Turbine Technicians require 10 keystone skills. Machinists already hold 5. The five-skill gap:

The two highest-priority gaps (Management of Material Resources and Instructing) both score 100 on job relevance — non-negotiable for the target role. Multilimb Coordination may require hands-on equipment training rather than classroom instruction. The workforce board now has a concrete, five-skill training design with clear priorities.

“Machinist → Wind Turbine Technician” sounds like a career change. At the skill level, it is 88% overlap with five trainable gaps, in a sector growing “much faster than average” with a 6.9-point resilience gain and a clean energy narrative that strengthens the case for funding.

The challenge: A consulting firm is writing a sector disruption report for a client in financial services. The client wants to know how AI exposure compares across industries and where their workforce sits relative to the broader economy. Every competing product offers sector-level averages. The firm needs to show the client why that approach is misleading.

Sector-Level Analysis Hides More Than It Reveals

Cubit scores every occupation individually, but we can aggregate by sector to show what sector-level reports actually look like — and what they miss. Using data aligned to nine GDPval economic sectors, Cubit automation scores produce a clear ranking:

The critical insight: The mean within-sector standard deviation (2.6) exceeds the between-sector standard deviation (2.0). Within-sector variation is larger than between-sector variation. This means sector-level analysis obscures more variation than it explains.

A Software Developer and an Architect both work in “Professional, Scientific, and Technical Services.” Their sector average is 45.5. But the range within that sector spans 42.1 to 50.4 — an 8.3-point spread that encompasses roles with fundamentally different exposure profiles. A sector report assigning both the same risk label would mischaracterize both.

The Information Sector: Highest Internal Variation

The Information sector is particularly revealing. Its mean score (44.3) places it mid-table, but its within-sector SD is 4.0 — the highest of any sector, producing a range from 40.2 to 50.1. A sector report calling Information “moderately exposed” would apply equally to the most-exposed and least-exposed occupations in the group. Cubit resolves this by scoring every occupation individually and then letting analysts aggregate at whatever level is meaningful for their context.

For the Client's Report

The consulting firm's deliverable is transformed. Instead of the standard “Finance ranks 4th out of 9 sectors on AI exposure” slide, they can show:

• Finance & Insurance has a tight within-sector SD of 1.6 — the second lowest. The sector is internally homogeneous on exposure.
• But individual roles still span 41.9 (Personal Financial Advisors) to 46.4 (Accountants). Custom scoring for client-trust weight amplifies that gap dramatically (see Enterprise demo above).
• Nine entire sectors — Construction, Education, Transportation, Installation & Maintenance, among others — have zero representation in competing datasets. These account for roughly half of US GDP.
• Cubit covers all 923 US occupations across all 22 SOC major groups. No gaps. No sampling bias.

Sector-level reports are the norm. They are also the floor. Cubit doesn't replace sector analysis — it reveals when sector analysis is enough and when it is not. For a consulting firm, the ability to quantify within-sector heterogeneity is a differentiated deliverable that no competitor can replicate from static datasets.

These demonstrations share a structural lesson: the gap between “AI will transform work” and actionable intelligence has been a data problem, not a conceptual one.

None of the existing alternatives deliver these analyses. GDPval provides binary evaluations for 44 occupations with no regional data, custom scoring, transitions, or skill gap analysis. The ILO scores 427 occupations on a single dimension with no task decomposition or API access. The CAIS Remote Labor Index evaluates 10 freelance projects with no occupational taxonomy. Each answers a narrow research question; none answers the operational questions decision-makers face daily.

Cubit closes each gap with the same architecture: task-level scoring across four dimensions, rolled up to occupation and region, and exposed through an API designed for integration. This enables macro-level policy, meso-level corporate strategy, and micro-level career planning — all from the same dataset, all through the same API, and all with real data.