There is a chart buried on page nine of a Barclays Research report that stopped me in my tracks. It shows the evolution of what global C-suite executives consider the most critical skills for their workforce, tracked across three survey waves: 2016, 2018, and 2023. In 2016, Proficiency in STEM sat comfortably at the top of the ranking, crowned at 42%. By 2023, it had plummeted to the very bottom of the list, at 28%. Meanwhile, the skill that climbed from relative obscurity to claim the throne was not machine learning, not prompt engineering, not data science. It was time management and the ability to prioritize.
Let that sink in for a moment. In the middle of the most explosive technological revolution since the personal computer, the world’s most powerful executives are not crying out for more engineers. They are crying out for people who know how to think clearly about what matters.
Source: IBM Institute for Business Value, Barclays Research. Figure 6 from “AI Revolution: Productivity Boom and Beyond,” Impact Series 12, January 2024.
The Report That Connects the Dots
The chart comes from “AI Revolution: Productivity Boom and Beyond”, the twelfth installment of Barclays’ Impact Series, published in January 2024 in partnership with the IBM Institute for Business Value . It is not your typical breathless AI hype piece. Instead, it takes a deliberately macro-economic lens, asking a deceptively simple question: can AI deliver the kind of productivity surge that the steam engine, electrification, and the personal computer each triggered in their time?
The answer, according to the authors, is a cautious but substantiated yes. But the path to that productivity boom is far more nuanced than simply deploying more algorithms. It requires a fundamental rethinking of how work gets done, what skills matter, and what role humans play in an increasingly automated landscape.
The report draws on two massive IBM IBV studies: one surveying 3,000 global C-suite executives across 28 countries and 20 industries, and another polling 21,000 workers across 22 countries . The scale of the data alone makes it one of the most comprehensive snapshots of executive sentiment on AI and workforce transformation available today.
Three Centuries of Productivity Waves, and the One We Are Waiting For
One of the most striking visualizations in the report traces the effect of technological breakthroughs on labour productivity from 1760 to the present day. The steam engine, the light bulb, the personal computer: each invention triggered a measurable surge in output-per-hour-worked, though often with a significant lag. James Watt patented his steam engine in the 1760s, but the peak productivity gains did not materialize for nearly a century.
Source: Kendrick (1961), Syverson (2013), BLS, BoE, Barclays Research. Figure 2 from “AI Revolution: Productivity Boom and Beyond.”
The implication is both hopeful and sobering. AI has two core attributes that previous revolutions lacked: accessibility (anyone can issue instructions to ChatGPT without learning a programming language) and versatility (the same LLM that summarizes a legal document can summarize a medical one, or generate marketing copy, or analyze financial data). These twin advantages suggest that AI could encounter fewer adoption obstacles than past technologies, potentially compressing the timeline from invention to macroeconomic impact.
Yet the productivity puzzle of the last two decades looms large. Why did smartphones and ubiquitous internet access not deliver the productivity boom economists expected? The report offers three possible explanations: the typical long lag between invention and macro gains, the possibility that digital tools simply do not match the transformative power of electricity or the automobile, and measurement errors in how GDP captures the value of zero-marginal-cost digital products. The truth, as the authors wisely note, likely lies in a combination of all three.
The Productivity Gap AI Needs to Fill
But here is where the report gets truly urgent. The authors overlay demographic projections from the United Nations onto historical productivity data for six major economies, and the picture that emerges is stark. As working-age populations shrink across the developed world, these countries need a dramatic acceleration in labour productivity just to maintain the modest growth rates they achieved before COVID. Italy would need a nearly 2.5 percentage point increase in annual productivity growth, a level significantly higher than anything the country has achieved in any five-year period since 1990. Germany and Japan each require boosts of 1 to 1.25 percentage points. Even the relatively younger economies of the US, UK, and France face substantial gaps.
Source: OECD, Barclays Research. Data elaborated from Figure 3 of “AI Revolution: Productivity Boom and Beyond,” Impact Series 12, January 2024.
This is the fourth wave the chart on the previous section implicitly points to. The steam engine, electrification, and the PC each delivered their productivity dividends over decades. AI is being asked to deliver its dividend faster, not because the technology demands it, but because the demographics do. The question is no longer whether AI can boost productivity. It is whether it can do so quickly enough to offset the structural headwinds of ageing societies. And the report suggests that the answer depends less on the technology itself than on how organizations choose to deploy it.
Augmented, Not Replaced: The 87% Consensus
Perhaps the most reassuring finding for anyone anxious about their job security is this: 87% of C-suite executives surveyed believe their employees are more likely to be augmented than replaced by generative AI . This is not a marginal majority. It is an overwhelming consensus that the future of AI in the enterprise is about human-machine partnership, not human-machine substitution.
The data becomes even more interesting when broken down by function. Procurement leads the augmentation expectation at 97%, followed by risk and compliance at 95% and finance at 93%. The functions where replacement is considered more likely, relatively speaking, are marketing (73% augmented) and customer service (77% augmented), though even here the vast majority of executives see augmentation as the dominant outcome.
Source: IBM Institute for Business Value, Barclays Research. Figure 7 from “AI Revolution: Productivity Boom and Beyond.”
“AI has the potential to transform the employee experience. It can automate repetitive tasks, letting people focus on what they are passionate about, freeing up their time for skills development or work-life balance, and potentially create exciting new job roles and career paths.” Jill Goldstein, Managing Partner, IBM Talent Transformation Consulting
This is the crux of the augmented workforce thesis. When AI handles the repetitive, the routine, and the computational, humans are freed to do what machines still cannot: exercise judgment, build relationships, navigate ambiguity, and lead with empathy. The skills that matter in this new division of labor are not the ones you learn in a coding bootcamp. They are the ones you develop through experience, self-awareness, and deliberate practice.
The Great Skills Inversion
Which brings us back to that extraordinary chart. The full ranking of most critical workforce skills in 2023, as identified by global executives, tells a story of profound inversion:
| Rank | Skill | 2016 | 2023 | Trend |
| 1 | Time management skills and ability to prioritize | 33% | 42% | Rising |
| 2 | Ability to work effectively in team environments | 35% | 40% | Rising |
| 3 | Ability to communicate effectively | 38% | 38% | Stable |
| 4 | Willingness to be flexible, agile, adaptable to change | 38% | 38% | Stable |
| 5 | Analytics skills with business acumen | 32% | 35% | Rising |
| 6 | Ethics and integrity | 31% | 33% | Rising |
| 7 | Industry/occupation specific skills | 31% | 33% | Rising |
| 8 | Proficiency in reading, writing, and mathematics | 33% | 32% | Stable |
| 9 | Foreign language | 28% | 32% | Rising |
| 10 | Capacity for innovation and creativity | 32% | 31% | Stable |
| 11 | Basic computer and software application skills | 40% | 31% | Falling |
| 12 | Proficiency in STEM | 42% | 28% | Falling |
The pattern is unmistakable. The two skills that have fallen most dramatically are basic computer skills (from 40% to 31%) and STEM proficiency (from 42% to 28%). These are precisely the domains where AI excels and where it is rapidly commoditizing human expertise. When a machine can write code, analyze datasets, and build models faster than most humans, the premium on those skills as standalone capabilities diminishes.
What rises in their place is a constellation of distinctly human competencies: the ability to manage time and prioritize in a world of infinite information, the capacity to collaborate effectively in increasingly distributed teams, and the ethical judgment to navigate the complex moral terrain that AI introduces. The report puts it elegantly: AI can complement a person’s particular blend of attributes, but it can hardly replace them.
This is not to say that STEM skills are becoming irrelevant. The report is careful to note that increasing use of AI is likely to change how those hard skills are deployed within organizations, and that a premium will attach to employees who can offer a combination of hard and soft skills. The pure technologist who cannot communicate, prioritize, or work in a team is becoming a liability. The hybrid professional who blends technical fluency with human intelligence is becoming the most valuable asset in the enterprise.
The J-Curve of AI Returns
For executives weighing the business case for AI investment, the report offers a sobering but ultimately optimistic picture. Average ROI on traditional AI projects climbed from roughly 1% in early 2020 to about 6% by the end of 2021. With the addition of generative AI, projections suggest ROI could reach approximately 10% by 2025 .
The pattern follows what the authors call the “J-curve”: a slow start as organizations work out the kinks, followed by an exponential rise once a critical threshold of adoption and integration is reached. Best-in-class AI performers, those who have developed mature capabilities across six key areas (vision and strategy, AI operating model, AI engineering and operations, data and technology, talent and skills, and culture and adoption), reported an average ROI of 13% on AI projects.
Critically, the report finds that AI adopters who also invest heavily in reskilling their workforce see a 36% revenue growth rate premium over other AI adopters. The technology alone is not enough. It is the combination of AI deployment with human capital development that unlocks the full value.
2025 Update: From Theory to Proof (and a $2.9 Trillion Question)
The Barclays-IBM report was published in January 2024, at a moment when generative AI was still more promise than proof. Nearly two years later, the data is starting to come in, and it both validates and complicates the original thesis.
McKinsey Global Institute’s November 2025 report, “Agents, Robots, and Us: Skill Partnerships in the Age of AI” , offers the most comprehensive update available. After analyzing over 11 million US job postings and 2,000 work activities across 800+ occupations, the findings are striking. Currently demonstrated technologies could theoretically automate activities accounting for about 57% of US work hours. But this is not a forecast of job losses. It is a measure of how profoundly work will change.
The most telling data point: more than 70% of the skills sought by employers today are used in both automatable and non-automatable work. Skills are not becoming obsolete. They are being redeployed. Workers will spend less time preparing documents and doing basic research, and more time framing questions and interpreting results. The human role is shifting from executor to orchestrator, a transformation that McKinsey argues could unlock $2.9 trillion in annual economic value in the United States alone by 2030.
| Metric | Barclays/IBM (Jan 2024) | McKinsey (Nov 2025) |
| Core thesis | Augmentation over replacement | Skill partnerships between people, agents, and robots |
| Executive consensus on augmentation | 87% of C-suite | 70%+ of skills remain relevant across automatable and non-automatable work |
| Fastest-growing skill demand | Time management and prioritization | AI fluency (7x growth in 2 years) |
| Fastest-declining skill demand | STEM proficiency (42% to 28%) | Routine writing and research |
| Economic value projection | 36% revenue premium for reskilling AI adopters | $2.9 trillion annual value by 2030 (US only) |
| Key insight on workflows | “Automating bad processes won’t make them better” | Maximum value comes from redesigning entire workflows, not individual tasks |
Perhaps the most explosive single data point comes not from McKinsey but from the MIT Media Lab, cited in a September 2025 Harvard Business Review article : 95% of organizations see no measurable return on their AI investments. This is not a contradiction of the optimistic projections. It is a confirmation of the Barclays report’s central warning. The organizations that are failing are the ones that simply bolted AI onto existing processes. The ones succeeding are those that redesigned their workflows around human-AI collaboration, investing in what McKinsey calls “skill partnerships” between people, agents, and robots.
The concept of “AI fluency”, the ability to use and manage AI tools effectively, has emerged as the breakout skill of 2025. Demand for it in US job postings has grown nearly sevenfold in just two years , faster than for any other skill tracked. This is not a niche technical requirement. It is showing up across management, finance, marketing, and education. It is, in many ways, the practical manifestation of the “time management and ability to prioritize” that executives identified as the top skill in the 2023 IBM survey. AI fluency is not about knowing how to code a model. It is about knowing when to use AI, when to override it, and how to direct it toward outcomes that matter.
McKinsey’s new Skill Change Index offers a granular view of which skills face the most disruption by 2030. The pattern mirrors and extends the Barclays findings: highly specialized, automatable skills like accounting, invoicing, and SQL face the greatest exposure, while interpersonal skills such as negotiation, coaching, and leadership are likely to change the least. The vast middle ground, including problem-solving, communication, and quality assurance, will evolve as part of the growing partnership with intelligent machines.
What This Means for Leaders, Builders, and the Rest of Us
The implications of these reports extend far beyond the C-suite. For anyone building a career, managing a team, or leading an organization, the message is clear: the age of the augmented workforce demands a new kind of professional. Not one who competes with machines on computational tasks, but one who leverages machines to amplify uniquely human capabilities.
The organizations that will thrive are not those that simply automate existing processes. As the Barclays report warns, “automating bad processes won’t make them better.” McKinsey’s 2025 research puts a number on it: after analyzing 190 business processes, they found that maximum value comes when AI agents act as “virtual coworkers” embedded in redesigned workflows, projecting a 7-12% revenue uplift for organizations that get this right . The winners will be those that strip their operating models down to the studs, reimagine workflows around human-AI collaboration, and invest in the soft infrastructure of skills, culture, and ethical governance that makes augmentation possible.
For the global economy, the stakes are enormous. Ageing workforces in developed countries and low per-capita output in developing nations represent two of the most pressing structural challenges of our time. AI, deployed as a complement to human labor rather than a substitute, could help address both. But only if we get the policy mix right, both at the regulatory level and within the enterprise.
The steam engine took a century to deliver its full productivity dividend. The light bulb took decades. The personal computer took a generation. AI, with its unprecedented accessibility and versatility, might just compress that timeline dramatically. But the catalyst will not be the technology itself. It will be the humans who learn to wield it wisely, prioritize what matters, and work together to build something greater than any algorithm could produce alone.
The most valuable skill in the age of AI, it turns out, is knowing what to do next.
