Editorial
Since ChatGPT’s public release in late 2022, a familiar question has hung in the air: will generative AI (GenAI) actually unlock a new era of productivity? Three years on, the evidence is mounting, and executives are paying attention.
The signals are striking. McKinsey’s Superagency in the Workplace report found that 94% of employees and 99% of C-suite leaders are at least familiar with GenAI tools, but leaders still underestimate actual adoption. They assume only 4% of staff are using AI for more than a third of their work, while employee self-reports suggest it is closer to 12%.
McKinsey estimates that GenAI could increase labor productivity by 0.1% to 0.6% annually through 2040, and the Penn Wharton Budget Model projects that AI adoption could lift U.S. productivity growth by 1.5% by 2035.
Yet the picture is far from simple. Productivity growth at the national level remains patchy, and the much-debated “productivity paradox” persists. Output gains are visible in pockets, but not evenly across economies or firms. The World Economic Forum warns of a widening productivity pay gap, with gains accruing to shareholders and the elite few at the top while wages for the median worker stagnate.
The real story lies not in aggregate numbers, but in the lived experience of workplaces where AI is being integrated.
From Call Centers to Code: What the Evidence Says
Three large-scale studies offer some of the clearest insights yet.
| Study | Context | Key Insights |
|---|---|---|
| Brynjolfsson, Li & Raymond (2025) – Generative AI at Work | Customer-support agents with AI assistants | Agents became 15% more productive overall, mainly by resolving issues faster. Newer agents improved the most, increasing their output by about 30%, and the gains lasted even when the AI tool was temporarily unavailable, suggesting that employees learned and internalized better ways of working with the help of AI. The most experienced and highest-skilled workers saw small gains in speed and small declines in quality. |
| Dell’Acqua et al. (2023) – Jagged Technological Frontier | Knowledge workers (consultants) using GPT-4 | Consultants finished 12% more tasks, worked 25% faster and work quality improved by over 40%. Lower-performing consultants saw the biggest gains, closing the gap with their higher-performing peers. When they used AI for problems that were too complex or outside its strengths, accuracy dropped. |
| Cui et al. (2025) – Software Developers & Generative AI | High-skilled software engineers using coding assistants | Developers completed 26% more coding tasks and increased their build activity by 13–38%, depending on the type of work. Junior developers benefited far more than senior ones, boosting their productivity by up to 39%, while experienced developers saw smaller gains of 8–13%. Code quality, measured by approval rates, improved by around 10%, though results varied by company. |
The findings echo across industries. Customer support agents armed with AI assistants handled calls faster and with fewer escalations. Consultants produced more and better work, particularly those in the bottom half of performance. Software developers with GitHub Copilot not only wrote more code but also saw higher approval rates, a proxy for quality.
In all three settings, the largest beneficiaries were not the most experienced or elite performers, but the novices, the strivers and the merely average.
Generative AI: The Great Leveler and Its Limits
The studies converge on a surprising pattern: AI is not a skill-biased technology in the traditional sense. Unlike earlier waves of automation, which tended to reward the highly educated and leave the less skilled behind, GenAI disproportionately lifts novices and compresses performance gaps.
Brynjolfsson et al. found that new agents effectively gained six months’ worth of learning in just two months with AI assistance. Dell’Acqua and colleagues documented a striking 43% productivity surge among bottom-half performers in consulting tasks. Cui et al. showed that short-tenure and junior developers increased their output by nearly a third, while senior engineers saw modest single-digit improvements.
But the “great leveler” effect has limits. The same studies found that top performers sometimes stagnated, or even saw declines, when relying heavily on AI. Customer-support veterans occasionally produced lower-quality conversations with AI. Consultants misapplied AI outside its competence frontier and were 19 percentage points less likely to be correct. Developers risked falling into a “trial-and-error” loop, compiling code more often without necessarily producing better solutions.
In short, AI lifts the floor but may lower the ceiling. The frontier of human judgment, creativity and innovation cannot simply be outsourced to algorithms.
Leading the Productivity Revolution
The takeaway for leaders should be that the productivity windfall is real, but requires deliberate steering. AI is not a plug-and-play tool; it is a system-level shift that reshapes learning, workflows and even organizational design.
1. Treat AI as a Capability Frontier, Not a Magic Wand
Executives should map where AI excels, i.e, tasks that are frequent enough to train models but not so trivial that humans already handle them well. These “moderately rare” tasks are the sweet spot. Beyond that frontier, AI erodes value.
2. Reimagine Workforce Development
AI’s greatest gift may be its role as a teacher. Novices absorb best practices faster, junior developers learn coding patterns from Copilot, and call-center staff pick up communication strategies. Leaders must double down on this “AI as a coach” phenomenon by redesigning onboarding, pairing personalized AI assistance with coaching, and ensuring tacit knowledge is not lost.
3. Safeguard Expertise
The risk for top performers is complacency. Firms must encourage experts to push boundaries rather than lean too heavily on AI-generated defaults. Incentives should reward originality, problem framing and cross-disciplinary insight (the human edge AI cannot replicate). Progressive leaders I work with are already thinking about how they can leverage AI to capture organizational wisdom (e.g., when a seasoned expert retires or a top performer resigns).
4. Redesign Workflows, Not Just Tools
Firms that merely hand out AI licenses will squander the opportunity. McKinsey’s survey shows that companies redesigning workflows and assigning C-suite oversight are the ones capturing measurable financial impact. This means rethinking process flows, team structures and performance metrics.
5. Manage the Productivity–Pay Divide
If productivity gains are captured only by shareholders and top executives, resentment will grow. Leaders must ensure that AI-driven efficiency translates into shared prosperity, whether through wage growth, career development or reinvestment in innovation.
6. Provide Guardrails for Responsible Use
As Dell’Acqua’s “jagged frontier” study shows, AI fails spectacularly outside its domain. Organizations need robust guardrails; this means training employees in critical evaluation, building escalation protocols and investing in AI governance structures.
Reimagine Work, Not Just Tools
GenAI is not the first technology to promise a productivity revolution. Electricity, the personal computer and the internet each produced bursts of optimism, disillusionment and eventual transformation. The difference today is speed; adoption cycles once measured in decades are now compressed into years or months.
The studies from Brynjolfsson, Dell’Acqua and Cui suggest that AI will not replace work so much as it will reshape who thrives, how fast people learn and how organizations must adapt. For executives, the opportunity is profound, but so is the responsibility.
The gains are there for the taking: faster throughput, improved quality and accelerated learning. However, the risks of deskilling, misapplication and widening inequality are equally real. To capture the promise, leaders must go beyond tool adoption and reimagine the very fabric of work.
Conference
Oct
20
Gartner IT Symposium/Xpo Orlando 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
The productivity paradox may not disappear overnight, but the contours of the future are becoming clearer. GenAI is already redefining work. Whether this shift translates into lasting gains or fleeting hype will hinge on the actions leaders take now.
Editor's Note: What else should leaders be considering as they roll AI out to their workforce?
- Metacognition: Your AI Productivity Edge — AI boosts creativity when paired with metacognition — reflecting on your thinking. This self-awareness drives learning velocity over mindless productivity.
- Lead or Get Run Over: A CEO's Field Guide to AI You Can Use Now — Real AI leadership means clarity over chaos. Start small, educate everyone, run ethical pilots and scale what works. Show proof, not PowerPoints.
- People-Centric Leadership and AI: Can We Have Both? — AI can support people-first leadership, if we use it with care.
Learn how you can join our contributor community.
