Work Map
Methodology

How the Co-Mind Work Map scores tasks

The Work Map classifies HR and Sales tasks into five collaboration zones based on where human judgment, AI capability, or their deliberate combination creates the most value. Here's how we score, why, and what research informs the model.

← Back to the Work Map

The Five Scoring Dimensions

Every task is scored on five dimensions, each rated 1–10. These dimensions capture the characteristics that research shows predict whether a task benefits most from human effort, AI capability, or deliberate collaboration between the two.

COG

Cognitive Load

How much analysis, synthesis, and pattern recognition does the task require? Tasks with high cognitive load involve processing complex information, connecting disparate data, and drawing non-obvious conclusions.

1 = Simple
10 = Heavy analysis
CMP

Complexity

How many steps, dependencies, and contextual variables are involved? Complex tasks require navigating ambiguity, managing trade-offs, and making judgment calls where the "right answer" depends on context.

1 = Straightforward
10 = Multi-layered
CRE

Creativity

How much novel thinking, ideation, or design does the task demand? Creative tasks produce original outputs — strategies, programmes, communications — rather than following established templates or rules.

1 = Template-based
10 = Original creation
RTN

Routine Level

How repetitive and predictable is the task? High-routine tasks follow consistent rules, produce similar outputs each time, and don't require judgment to handle variations. This is the strongest predictor of automation potential.

1 = Every time is different
10 = Highly repetitive
SOC

Social & Empathy

How much interpersonal interaction, trust-building, and emotional intelligence does the task require? Tasks scoring high on social demand human presence, reading between the lines, and navigating sensitive dynamics.

1 = No human interaction
10 = Deeply interpersonal

The Anthropic Economic Index found that social intelligence showed near-zero correlation with AI usage across all five of its parameters, positioning it as a durable human comparative advantage. Meanwhile, high cognitive load and complexity are the strongest signals for productive human–AI collaboration.

The Five Collaboration Zones

Based on its dimension scores, each task maps to one of five collaboration zones. These zones form a spectrum from fully human to fully automated — with Augment at the centre, where deliberate collaboration design creates the most value.

Human Essential

Tasks where human judgment, empathy, and presence are irreplaceable. AI has minimal or no role. These typically involve high-stakes interpersonal dynamics, ethical decisions, or situations requiring trust and confidentiality.

Signal: High SOC (≥8) + Low RTN (≤3)

Human-Led, AI Assists

The human drives the task, but AI provides supporting inputs — data preparation, research, option generation. The human makes all decisions, shapes the output, and manages stakeholder relationships.

Signal: High CMP (≥6) + High SOC (≥5) + Moderate COG

Augment

The sweet spot. Tasks where neither human nor AI alone produces the best outcome, but their deliberate collaboration does. AI handles pattern recognition, data processing, and first drafts; the human provides context, judgment, and refinement. This is where CoMindLab's approach creates the most value.

Signal: High COG (≥7) + High CMP (≥6) + Low RTN (≤5) + Moderate SOC

AI-Led, Human Oversight

AI does the heavy lifting — data processing, report generation, pattern scanning. A human reviews outputs, handles exceptions, and intervenes when the AI encounters edge cases or ambiguity.

Signal: High RTN (≥6) + Low SOC (≤4) + Moderate COG

Fully Automated

Rules-based, high-volume tasks with predictable inputs and outputs. No human judgment needed. These are the "set and forget" processes — scheduling, form processing, standard calculations.

Signal: High RTN (≥8) + Low SOC (≤2) + Low CMP (≤4)

Zone Classification Logic

The zone is determined by the dominant pattern across all five dimensions. Here's the decision matrix, applied in priority order:

Zone Primary Signals Example Task
Human Essential SOC ≥ 8 and RTN ≤ 3 Conducting workplace investigations
Human-Led SOC ≥ 5, CMP ≥ 6, RTN ≤ 4 Strategic workforce planning with leaders
Augment COG ≥ 7, CMP ≥ 6, RTN ≤ 5, SOC < 8 Pay equity analysis
AI-Led RTN ≥ 6, SOC ≤ 4 Salary benchmarking & market analysis
Automated RTN ≥ 8, SOC ≤ 2, CMP ≤ 4 Payroll processing & reconciliation

Important nuance: These are guidelines, not rigid rules. Some tasks have been manually classified where the dimension scores alone don't capture the full picture. For example, "ethics review of analytics models" scores high on COG (9) but is classified as Human Essential because the judgment required is fundamentally about values and fairness, not pattern recognition.

Research Foundations

The scoring model draws on multiple research streams, each contributing a different lens on how human and AI capabilities intersect at task level. Browse all 28 research sources with key excerpts →

01
Anthropic Economic Index (2025) — Privacy-preserving analysis of millions of real Claude conversations mapped to 20,000+ O*NET tasks. Provides observed AI usage rates for specific HR tasks and the 7-dimension task scoring framework (cognitive load, complexity, creativity, decision-making, routine, social intelligence, domain knowledge). Our 5-dimension model is a simplified adaptation. Key finding: 57% of occupations show significant AI usage for at least one task, but only 4% are >75% automatable. Read on anthropic.com → Where to find it: Section 3 "Task-Level Analysis" and Appendix B for dimension definitions
02
Dell'Acqua, F. et al. "Navigating the Jagged Technological Frontier" (2023) — BCG / Harvard Business School. Preregistered experiment with 758 BCG consultants on 18 realistic tasks. Found 40% quality improvement for tasks within the AI frontier, but 19% worse outcomes when AI was applied to tasks outside it. Established the Centaur (strategic task division) and Cyborg (deep interweaving) collaboration archetypes. Read on SSRN → Where to find it: pp.12–18 for the Centaur/Cyborg framework; Table 3 for task-by-task results
03
Stanford HAI "WORKBank" (Terzis et al., 2025) — Surveyed 1,500 domain workers and 100+ AI experts across 844 tasks in 104 occupations. Created the Human Agency Scale (H1–H5) showing 45.2% of occupations prefer equal human–AI partnership. Our five collaboration zones directly map to this H1–H5 scale. Read the paper on arXiv → Where to find it: Figure 4 for the Human Agency Scale distribution; Appendix C for task-level classifications
04
McKinsey & Company "Generative AI and the Future of HR" (2024) — Estimated two-thirds of HR tasks can be automated to a "large degree," with employees spending 60–70% less time on administrative work. Provided function-level value distribution: Talent Acquisition (20%), People Management (20%), and Org Planning (15%). Read on mckinsey.com → Where to find it: Exhibit 2 for function-level automation estimates; "The HR value chain" section for task breakdowns
05
Mercer "Three HR Roles Transformed by Generative AI" (2024) — Detailed task-time analysis for HRBPs, L&D specialists, and Total Rewards leaders. Found AI reduces HRBP talent management time from 55% to 37%, L&D programme design time from 35% to 21%, and that 52% of Total Rewards workload is AI-affectable. Read on mercer.com → Where to find it: Role-by-role time-allocation charts; "HRBP" section for the 55%→37% finding
06
SHRM "2025 Talent Trends Report" — AI adoption in HR climbed to 43% in 2025 (from 26% in 2024). Recruiting leads at 51% adoption. 66% of AI-using recruiters use it for writing job descriptions; 44% for screening. Provided adoption benchmarks by HR function. Read on shrm.org → Where to find it: Chapter "AI in Talent Acquisition" for adoption %; Figure 12 for AI usage by task type
07
Gartner "Top 5 HR Trends and Priorities for 2026" — 82% of HR leaders plan to use agentic AI by May 2026. 88% say their organisations have not realised significant business value from current AI tools. Employees in AI-relevant roles save an average of 1.5 hours per day. Read on gartner.com → Where to find it: "Agentic AI" trend section for the 82% finding; "AI value realisation" for the 88% figure
08
Deloitte "2025 Global Human Capital Trends" — 59% of organisations take a technology-focused approach to AI; they are 1.6x more likely to NOT realise returns versus human-centric approaches. 83% of companies report low workforce analytics maturity. Introduced the "boundaryless HR" framework. Read on deloitte.com → Where to find it: Chapter 3 "Digital playground" for the 1.6x finding; "Human performance" section for analytics maturity data
09
Eloundou, T. et al. "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models" (2024) — Published in Science. Foundational task exposure methodology mapping LLM capabilities against O*NET task descriptions. Found ~80% of the US workforce could have at least 10% of tasks affected; ~19% could see 50%+ of tasks affected. Read in Science → Where to find it: Table 1 for exposure estimates by occupation; Methodology section for the task-level rubric we adapted
10
HR Acuity "The State of Employee Relations & AI" (2025) — Only 1% of organisations widely use AI in employee relations; 35% are experimenting. Identified investigations, mediation, and disciplinary proceedings as the irreplaceable human core, with AI value limited to documentation, pattern detection, and compliance scanning. Read on hracuity.com → Where to find it: Section "AI in ER" for the 1%/35% adoption figures; "Use Case Matrix" for task-level recommendations
11
World Economic Forum "Future of Jobs Report 2025" — Projects human-machine collaborative tasks growing from 30% to 33% by 2030, while human-only tasks decline from 47% to 33%. 77% of employers plan to prioritise reskilling for AI collaboration. Surveyed 1,000+ employers across 22 industries. Read on weforum.org → Where to find it: Chapter 4 "Technology" for task share projections; Figure 4.1 for the 30%→33% chart; Appendix B for industry breakdowns

How This Compares to Other Frameworks

Several major frameworks classify work tasks against AI capability. Here's how the Co-Mind Work Map relates to each:

Framework Approach Our Relationship
Anthropic Economic Index 7 dimensions, observational (actual AI usage patterns) Our primary inspiration. We simplified 7 dimensions to 5 for usability, and added zone classification logic on top.
OpenAI "GPTs are GPTs" Task exposure scoring (0/0.5/1) against O*NET Binary exposure model. We extend this by measuring how AI should be involved, not just whether it can be.
Stanford HAI WORKBank Human Agency Scale (H1–H5), expert-rated Closest to our zones. Their H1–H5 maps directly to our Human Essential → Automated spectrum. We add dimension scoring for transparency.
BCG Jagged Frontier Inside/outside frontier classification Their frontier concept maps to our zone boundaries. We make the boundaries explicit with scoring rules instead of empirical task-by-task testing.
WEF Future of Jobs Macro projections (% of task hours by category) WEF projects at macro level; we apply at task level. Their 30%→33% collaborative task projection validates our Augment sizing.

What makes our model different: Most frameworks tell you whether AI can do a task. We focus on how humans and AI should work together on that task. The five zones aren't just a classification — they're a design specification for building the right collaboration pattern.

Limitations & Honest Caveats

This model is a starting point, not a definitive classification. Important caveats:

  • Tasks are generalised. The same task title can mean very different things in a 50-person startup versus a 50,000-person multinational. Context matters enormously.
  • Scores reflect current AI capability (2026). The frontier moves. Tasks that are "Human Essential" today may shift as AI capabilities evolve — and vice versa.
  • 8 tasks per role is a simplification. Most HR roles involve 20–40 distinct activities. We selected the 8 most representative to keep the tool usable.
  • Organisational maturity varies. An organisation with mature data infrastructure will see different AI potential than one still running on spreadsheets.
  • Regulatory environment matters. The EU AI Act classifies HR AI as high-risk. What's technically possible may not be legally permissible in all jurisdictions.
  • This is not a replacement for human judgment about your own work. You know your tasks better than any model. Use this as a conversation starter, not a verdict.

Ready to map your work?

Explore the Work Map to see how roles break down across collaboration zones — with task-level scoring and use case recommendations.

HR Edition Sales Edition HR Insights Report