The Task Exposure
OpenAI 'GPTs are GPTs' Study, August 2023
OpenAI researchers used O*NET to measure how LLMs could affect the U.S. labor market. Their finding: 80% of workers have at least 10% of their tasks exposed to LLMs. But here's the critical insight: LLMs alone only speed up 15% of tasks. When you build systems around them, that number jumps to 56%. The gap between 15% and 56% is the system integration work that most organizations skip.
Key Findings
LLMs exhibit traits of general-purpose technologies: broad impact across the economy, improvement over time, and significant co-invention. This is not a niche technology.
80% of workers have exposed tasks, but the magnitude varies. Higher-income jobs face greater exposure. Programming and writing skills are positively associated with LLM impact.
LLMs alone only speed up 15% of tasks. With complementary software and systems, that jumps to 47-56%. The gap is the system integration most organizations haven't done.
Science and critical thinking skills are negatively associated with exposure. The tasks that require human judgment, physical presence, and interpersonal skills remain human-owned.
80% of workers have exposed tasks, but LLMs alone speed up only 15%. The gap to 56% requires building systems around the models. That system integration is the work redesign most organizations skip.
Eloundou, Manning, Mishkin, Rock. 'GPTs are GPTs,' OpenAI, August 2023The Exposure Gap
Anthropic Nowcasting Report, March 2026
Anthropic measured AI task coverage across the U.S. economy using O*NET, the same occupational database Nuvepro uses. Their finding: 94% of occupations have at least one task AI can handle, but only 33% of tasks are actually being done by AI today. That 61-point gap is the exposure gap, and it exists because organizations deployed the technology without redesigning the work.
Key Findings
AI coverage is broad but shallow. Most occupations are exposed, but most tasks within those occupations aren't being handled by AI.
Programming and writing tasks lead in actual AI adoption. Physical, interpersonal, and judgment-heavy tasks lag behind.
30% of workers have zero AI task coverage. Their jobs haven't been touched by AI at all, even though the technology exists.
The gap between theoretical and actual coverage is the adoption problem. The tools are there. The work redesign isn't.
94% theoretical coverage vs. 33% actual. The 61-point gap isn't a technology problem. It's a work design problem. Organizations have the AI tools. They haven't redesigned the work to use them.
Anthropic, 'Nowcasting AI's Impact on the U.S. Labor Market,' March 2026The Training Gap
Guild Education AI Training Study, December 2023
Guild surveyed 355 workers and found a massive gap between AI tool deployment and workforce readiness. 88% of workers don't trust their employer to prepare them for AI. Only 36% of frontline employees have received any AI training. Meanwhile, demand for AI skills has grown 800% in 12 months. The tools are deployed. The people aren't ready.
Key Findings
88% of workers are not confident their employer will support them through the AI transition. Trust is broken before training even starts.
Only 36% of frontline employees say they've received any AI training, compared to 44% of leaders. The gap is widest for the people doing the work.
60% of workers believe they need new skills due to AI, but most employers have not taken a clear stance or provided guidelines for AI tool usage.
AI skills are concentrating at the top. Younger, affluent, educated workers in financial services have the highest AI usage. The equity gap is growing.
Workers know they need new skills. Employers aren't delivering. The trust gap and the training gap are two sides of the same problem: organizations deployed AI tools without a plan for the people.
Guild Education, 'AI Employee Training Framework,' December 2023 (survey of 355 members, June 2023)The Skills Gap
Everest Group L&D Report, January 2026
Everest Group surveyed enterprise L&D leaders and found the infrastructure for workforce transformation is missing. 47% of enterprises report skill shortage as their top challenge, but only 25% have virtual labs and only 18% have skills management platforms. The tools exist to train people. The systems to deliver that training at scale do not.
Key Findings
Skills-based learning is the #1 L&D priority, overtaking role-based training. Enterprises are shifting from training for job titles to training for specific capabilities.
Only 43% of enterprises have assessment tools for skill validation. Fewer than half can measure whether their training actually worked.
69% of enterprises outsource some or all digital learning delivery. Most organizations don't have the internal capability to build the training infrastructure AI demands.
38% of enterprises plan to increase L&D budget in the next year. Investment is growing, but it's being spent without the infrastructure to measure results.
Enterprise transformation is outpacing workforce preparedness. 47% say skills are the top gap, but only 25% have the labs to close it and only 43% can even measure progress. The L&D infrastructure itself is missing.
Everest Group, 'Getting the L&D Pulse: The State of Enterprise Learning in the Age of AI,' January 2026The Production Gap
Securing AI's Giant Leap, 2026
A cross-industry analysis of AI deployment outcomes reveals a devastating pattern: 80% of AI projects fail, 88% of pilots never reach production, and only 15% of companies have seen meaningful EBIT impact. The root cause: organizations spend 70% of their AI budget on models and tools, leaving almost nothing for data governance, AI security, and workforce preparation. The gap between excitement and production is where projects die.
Key Findings
The timeline from excitement to accountability is compressing. 2023: Excitement (GenAI wrappers). 2024: Engagement (POCs). 2025: Eagerness (Adoption Challenges). 2026: Expectations (ROI and Value Performance).
The difference between 'using AI' and 'doing AI': buying a tool for summarization yields marginal returns. Breaking functions into tasks and delegating to agents or human oversight unlocks exponential lifts.
70% of AI project timelines are spent on LLMs and tools. Production requires equal investment in data governance, AI security, and workforce preparation.
Replacement as a strategy yields 0-23% returns. Augmentation yields 200-500x. True convergence where AI and operations integrate seamlessly yields 2000x.
88% of AI pilots don't reach production. Not because the technology fails, but because organizations spend 70% of their budget on models and skip the work redesign, workforce preparation, and governance that production requires.
'Securing AI's Giant Leap: From Fragmented Hype to Exponential Human Amplification,' 2026
The GDP Impact
Cognizant / Oxford Economics, January 2024
Cognizant partnered with Oxford Economics to build an economic model analyzing generative AI's impact on 18,000 tasks across 1,000 jobs in the US economy. Their finding: gen AI could inject up to $1.043 trillion in annual GDP by 2032, but 90% of jobs will be disrupted. The difference between a productivity boom and a displacement crisis depends entirely on how organizations invest in their people.
Key Findings
Gen AI adoption follows an s-curve: 13% of businesses by 2026, 31% by 2030, 46% by 2033. The window for preparation is now.
CEO exposure score reaches 25%+ by 2032. General managers and ops managers: from 18% today to 52.7%. This is not just a frontline issue.
11% of displaced workers (about 1% of total workforce) may struggle to find new work. Left unmanaged, this creates structural unemployment.
The model uses 18,000 O*NET tasks across 1,000 jobs, the same task-level methodology that underpins Nuvepro's classification engine.
Adoption alone does not determine outcomes. Three scenarios show GDP impact ranging from $477B (low) to $1.043T (high) depending on how organizations invest in reskilling.
Same O*NET task-level methodology, same conclusion: 90% of jobs disrupted, but outcomes depend on reskilling investment. The $1T question is not whether AI changes work, but whether organizations prepare their people for the change.
Cognizant and Oxford Economics, 'New Work, New World with Generative AI,' January 2024The Displacement Gap
Goldman Sachs Research, August 2025
Goldman Sachs Research examined 800+ occupations to assess whether AI productivity gains translate into job displacement. Their finding: AI could displace 6-7% of the US workforce, but the impact is likely transitory. Historically, technology-driven displacement disappears after two years. The real concern is not long-term job loss but the transition period where displaced workers need new skills.
Key Findings
Technology-driven unemployment historically increases by 0.3 percentage points per 1% productivity gain, but this displacement disappears after two years.
85% of employment growth since 1940 has come from technology-driven job creation. The pattern of displacement followed by new roles is consistent.
AI adoption remains low: only 9.3% of companies report using gen AI in production. The labor market effects are still ahead of us, not behind us.
Sectors already contracting below trend: marketing consulting, graphic design, office administration, telephone call centers.
No significant statistical correlation yet between AI exposure and job growth, unemployment rates, or earnings growth. The transition has barely started.
Only 9.3% of companies use gen AI in production. When adoption scales, 6-7% of the workforce faces displacement. The two-year historical recovery window only works if reskilling infrastructure exists. Without it, transitory becomes permanent.
Goldman Sachs Research, Joseph Briggs and Sarah Dong, August 2025The Job Creation Gap
World Economic Forum Future of Jobs Report, February 2026
The WEF Future of Jobs Report projects 92 million jobs eliminated by 2030 but 170 million new roles created, yielding a net gain of 78 million. The gap between destruction and creation is not automatic. Organizations that invest in workforce development are 1.8x more likely to report better financial results. The framework requires five pillars: Vision, Skills, Technology, Process, and Culture.
Key Findings
Net positive: 170M new roles vs. 92M eliminated. But the 78M net gain requires intentional, inclusive workforce transformation.
Organizations investing in workforce development are 1.8x more likely to report better financial results (Deloitte 2025 Human Capital Trends).
Five-pillar framework: Vision (define future-ready), Skills (map capabilities), Technology (augment not replace), Process (redesign work), Culture (continuous learning).
Industry-specific transformation is essential: manufacturing needs smart factories, healthcare needs digitally enabled care teams, financial services need AI-augmented advisors.
Workforce transformation cannot be a standalone initiative or technology upgrade. It requires a systemic, industry-aware approach.
78M net new jobs is the prize, but only for organizations that invest in all five pillars. Technology alone is one pillar. The other four, Vision, Skills, Process, and Culture, are the work most organizations skip.
World Economic Forum, 'The AI-Driven Workforce Is Here,' February 2026The Productivity-Pay Gap
Economic Policy Institute, Updated March 2026
EPI's Productivity-Pay Tracker reveals the defining economic pattern of the last 45 years: productivity grew 92.4% while typical worker pay grew only 33.6%. The gap is entirely driven by rising inequality. From 1948 to the late 1970s, pay and productivity climbed together. Then policy choices broke the link. AI threatens to widen this gap further, unless the productivity gains reach workers.
Key Findings
Pay and productivity climbed together from 1948 to the late 1970s. The divergence was a policy choice, not an economic inevitability.
The entire gap is associated with rising inequality: inequality among wage earners and rising share of income going to capital owners rather than workers.
Class-based policies (tight labor markets, rising minimum wages, unionization) were so powerful they narrowed the Black-white earnings gap from the 1940s through 1970s.
AI could drive another 15% productivity surge (Goldman Sachs). Without worker-centric policies, the gap widens further.
The question is not whether AI boosts productivity, but who captures the value. Workers, shareholders, or both.
Productivity up 92%, pay up 34%. If AI drives the next productivity surge and workers don't capture the value, the gap becomes a chasm. Training workers to build and run the agents is how the value reaches them.
Economic Policy Institute, 'The Productivity-Pay Gap,' Updated March 23, 2026The AI Failure Gap
Pertama Partners AI Analysis, 2025
Pertama Partners analyzed enterprise AI project outcomes and found a devastating pattern: 80.3% of AI projects fail, 95% of GenAI pilots fail to scale beyond initial deployment, and the average sunk cost is $4.2 million per failed initiative. The root cause is not technology. 84% of failures are attributed to leadership gaps, organizational readiness, and missing workforce preparation.
Key Findings
95% of GenAI pilots fail to scale. The gap between pilot and production is not a technology problem.
84% of failures are attributed to leadership gaps, organizational readiness, and missing workforce preparation. The technology works. The organization doesn't.
$4.2M average sunk cost per failed initiative. Multiply by the number of pilots running across an enterprise, and the waste is staggering.
The failure rate has not improved despite massive increases in AI spending. More budget for models without investment in people produces the same result.
95% pilot failure rate at $4.2M each. The root cause is 84% leadership and readiness, not technology. Every dollar spent on models without matching investment in workforce preparation is likely wasted.
Pertama Partners, AI Failure Analysis, 2025
The Worker Productivity Gap
MIT Sloan & BLS Productivity Data, 2025-2026
MIT Sloan research shows organizations are 16% more productive when they listen to workers' input on AI implementation. Meanwhile, BLS data confirms the US economy is in a productivity acceleration: nonfarm business productivity grew 2.1% in 2025, above the prior cycle's 1.5%. The question is whether this acceleration reaches workers or concentrates at the top.
Key Findings
Organizations that involve workers in AI implementation decisions see 16% higher productivity than those that impose AI top-down.
US nonfarm business productivity grew 2.1% in 2025, the highest sustained rate since the 2007-2019 cycle (1.5%). The AI productivity wave may already be starting.
Labor share of output is 54.4% in Q4 2025. The remaining 45.6% goes to capital. AI could shift this ratio further.
Manufacturing productivity decreased 2.5% in Q4 2025 despite economy-wide gains. The benefits of AI are not evenly distributed across sectors.
16% more productive when workers have a say. The data is clear: top-down AI deployment underperforms. Train the people, involve them in the redesign, and productivity follows.
MIT Sloan Management Review (2025) & BLS Productivity and Costs, Q4 2025 Revised (March 2026)Why Task-Level Analysis
How research validates the approach
Anthropic uses O*NET to classify tasks as AI-coverable or not
Nuvepro uses O*NET to audit every task in your department and classify it as automate, augment, or human-only
The 61-point gap exists because work wasn't redesigned around AI capabilities
Nuvepro redesigns the work: which tasks agents own, which humans keep, where the handoffs happen
30% of workers have zero AI coverage despite available technology
Nuvepro identifies exactly which roles have untouched tasks that AI should handle
88% of workers don't trust their employer to prepare them for AI
Nuvepro trains the people hands-on until they're project-ready, not just 'AI aware'
AI skills concentrate at the top. Frontline workers are left behind
Nuvepro's Readiness Bundles work for all levels, from frontline workers to leadership
Only 25% of enterprises have virtual labs. 47% say skills are the top gap
Nuvepro provides GenAI Sandboxes, Simulations, and Readiness Bundles as the training infrastructure enterprises are missing
Only 43% of enterprises can assess whether training worked
EASE assessments validate that people are project-ready, not just 'trained'
LLMs alone speed up 15% of tasks. With systems, 56%. The gap is integration work
Nuvepro builds the system: auditing which tasks move to agents, designing handoffs, training the people
88% of AI pilots fail to reach production. 70% of budget goes to models, not workforce
Nuvepro is the workforce preparation layer that turns pilots into production
Cognizant analyzed 18,000 O*NET tasks and found 90% of jobs disrupted, with outcomes depending on reskilling
Nuvepro classifies the same 18,000+ tasks per occupation and builds the reskilling path for each one
Goldman Sachs: only 9.3% of companies use gen AI in production. The transition has barely started
Nuvepro's first-role-live-in-4-weeks approach gets organizations from 0% to production before the wave hits
WEF: 78M net new jobs, but only with intentional investment across Vision, Skills, Technology, Process, Culture
Nuvepro provides Skills (training), Technology (GenAI Sandboxes), and Process (task redesign) in one platform
EPI: productivity up 92%, pay up 34%. Workers don't automatically capture the value of technology gains
Nuvepro trains workers to build and run the agents, ensuring they capture the value of AI productivity
95% of GenAI pilots fail to scale. 84% of failures are leadership and readiness, not technology
Nuvepro's audit-first approach surfaces readiness gaps before pilots start, so the 84% failure mode is prevented
MIT Sloan: 16% more productive when organizations listen to workers on AI implementation
Nuvepro's task-level audit starts with what workers actually do, building AI integration from the work up