Why beginners may gain most from AI help

Introduction

One of the most striking findings from early workplace AI deployments is that beginners often gain more than experts. In several studies, the largest productivity improvements have appeared among newer, less experienced, or lower-performing workers rather than among top performers. That does not mean AI turns novices into masters overnight. It does suggest that AI assistants can compress part of the learning curve by providing guidance that previously depended on mentoring, accumulated experience, or access to particularly skilled colleagues.

This matters well beyond customer support centres. If advanced AI makes practical knowledge easier to access, one effect could be a reduction in the scarcity of expertise. Workers who start with less experience may spend less time stuck, less time waiting for help, and less time learning through costly trial and error. In the broader AI bloom vision, this raises a larger possibility: intelligence becoming more widely available as a service, allowing more people to participate in complex work, learning, and problem-solving. The evidence remains early and task-specific, but it offers a concrete example of how AI might distribute capabilities that were once concentrated in small groups.

What the customer support evidence shows

The strongest evidence comes from a widely discussed study by Erik Brynjolfsson, Danielle Li and Lindsey Raymond, who analysed the introduction of a generative AI assistant across more than 5,000 customer-support agents at a Fortune 500 software company. The system provided real-time suggestions during customer interactions. Researchers found that access to the assistant increased productivity by roughly 14–15% on average. More importantly, the gains were highly uneven. Novice and lower-skilled workers improved by about one-third, while experienced and highly skilled workers saw much smaller benefits.[OUP Academic]academic.oup.comOUP AcademicGenerative AI at Work* | The Quarterly Journal of Economicsby E Brynjolfsson · 2025 · Cited by 3397 — We study the effect of…[NBER]nber.orgGenerative AI at Workby E Brynjolfsson · 2023 · Cited by 3397 — In this paper, we study the staggered introduction of a generative AI…

The pattern was not simply a matter of working faster. The study found that less experienced workers often improved both speed and quality. Customer satisfaction remained stable or improved, and workers became better at handling difficult interactions. The researchers also found evidence that AI assistance improved English fluency and communication, especially for international employees.[arXiv]arxiv.orgarXiv Generative AI at WorkarXiv Generative AI at Work[Stanford Graduate]gsb.stanford.eduStanford Graduate School of BusinessGenerative AI at Work - Stanford Graduate School of BusinessWe also find evidence that AI assistance… School of Business

The central finding was that AI narrowed performance gaps. Instead of amplifying differences between the strongest and weakest workers, the tool appeared to reduce them. In economic terms, that is notable because many previous technologies have been described as “skill-biased”, meaning they reward workers who already possess scarce expertise. This deployment often looked more like skill diffusion.[SIEPR]siepr.stanford.eduGenerative AI at Work | Stanford Institute for Economic Policy…In this paper, we study the staggered introduction of a generative…[MIT Sloan]mitsloan.mit.edugenerative ai and worker productivityMIT SloanGenerative AI and Worker Productivity16 Apr 2024 — Access to the new tool increased worker productivity by about 14% on average…

How real-time suggestions shorten the learning curve

The most plausible explanation is not that the AI became smarter than expert workers. Rather, it helped beginners borrow expertise that had already been accumulated elsewhere in the organisation.

In traditional workplaces, much valuable knowledge is tacit. High performers often know how to phrase a difficult response, calm an angry customer, diagnose an unusual problem, or recognise subtle patterns, but they may struggle to explain exactly how they do it. New employees usually acquire those skills slowly through observation, feedback, and repeated mistakes.

Generative AI systems can sometimes extract patterns from huge numbers of past interactions and surface them in real time. Instead of waiting for a supervisor to review their work days later, a novice worker receives immediate suggestions while the task is happening. Brynjolfsson described this as capturing forms of tacit knowledge that had never been formally written down.[Time]time.comHow to Make AI Work for You, at WorkShe pursued this interest by taking the Elements of AI, an online course by MinnaLearn and the University of Helsinki, which enhanced her…

Several mechanisms help explain why the gains are concentrated among beginners:

The result is that beginners move down the experience curve more quickly. Researchers associated with the study explicitly argued that the technology helped newer workers acquire capabilities that would otherwise have taken longer to develop.[MIT Sloan]mitsloan.mit.edugenerative ai and worker productivityMIT SloanGenerative AI and Worker Productivity16 Apr 2024 — Access to the new tool increased worker productivity by about 14% on average…

Why experts often benefit less

The same mechanism that helps beginners can limit gains for experts.

If the AI has learned from historical examples, many of its recommendations may already resemble the methods used by top performers. An experienced worker therefore receives advice that is often close to what they would have done anyway. The gap between their existing performance and the AI’s suggestions is small.

In some cases, experts may even find generic recommendations mildly constraining. The customer-support study found small declines in quality among the highest-skilled workers despite modest speed gains. One interpretation is that experts possess contextual judgement that standardised suggestions cannot fully replicate.[arXiv]arxiv.orgarXiv Generative AI at WorkarXiv Generative AI at Work

This points to an important distinction. AI can help transfer established best practices, but expertise is not only the application of known rules. Experts often:

Those capabilities are harder to compress into a suggestion engine.

The finding therefore does not imply that expertise becomes irrelevant. It suggests that part of what organisations call expertise consists of accumulated procedural knowledge that can increasingly be shared rather than hoarded.

Where productivity gains stop short of expertise

A common misunderstanding is that narrowing performance gaps means eliminating them.

Most workplace studies measure outputs such as issues resolved per hour, task completion rates, coding performance, writing quality scores, or customer satisfaction metrics. These are useful indicators, but they are not the same as measuring deep professional competence.

A novice customer-support agent who receives AI guidance may perform more like a competent colleague during a particular interaction. That does not mean they understand the underlying systems as thoroughly. If the AI fails, the beginner may still lack the conceptual framework needed to solve unusual problems independently.

Researchers reviewing broader evidence on AI and work have noted that novice gains are most consistently observed in relatively structured tasks. As tasks become more complex, the evidence becomes less clear. Some studies still find larger gains among weaker workers, while others show benefits spread more evenly across skill levels.[arXiv]arxiv.orgarXiv Generative AI at WorkarXiv Generative AI at Work

This creates a tension that many organisations are beginning to confront. If AI makes workers more productive without requiring them to develop underlying expertise, companies may become more dependent on the tool itself. The immediate productivity gains are real, but long-term skill formation may become more complicated.

The risk of creating capable beginners but fewer masters

The optimistic interpretation is that AI democratises knowledge. The more cautious interpretation is that it may sometimes substitute for learning rather than accelerate it.

Historically, junior workers often learned by performing simpler versions of tasks later handled independently. If AI performs part of that scaffolding function, organisations may need new ways to ensure workers still develop deep understanding.

This concern appears in emerging research on AI and work transformation. Workers increasingly delegate routine elements of jobs to AI while taking on new responsibilities focused on checking, correcting, and managing outputs. That can improve efficiency, but it also changes how expertise develops.[arXiv]arxiv.orgarXiv Generative AI at WorkarXiv Generative AI at Work

The question is especially important in professions that rely on apprenticeship models. Law, software engineering, medicine, consulting, and scientific research all depend partly on newcomers learning from gradual exposure to increasingly difficult work. If AI handles much of the routine training ground, organisations may need deliberate strategies to prevent the erosion of future expertise.

The evidence today is too limited to resolve this question. Most studies track months rather than decades. They reveal how AI changes immediate performance, not how entire professions evolve.

Why this matters for a larger AI bloom future

The beginner-versus-expert pattern offers an early glimpse of a broader possibility in the AI bloom framework.

For most of human history, access to expertise has been constrained by geography, wealth, institutions, and the limited time of skilled people. Teachers can only teach so many students. Managers can only mentor so many employees. Specialists can only answer so many questions.

AI assistants hint at a world in which some forms of guidance become dramatically more abundant. The customer-support studies suggest that a portion of valuable know-how can be captured, reproduced, and delivered at scale.[NBER]nber.orgNBER WORKING PAPER SERIES GENERATIVE AI AT…by E Brynjolfsson · 2023 · Cited by 3306 — We find that access to AI assistance increases t…

If similar patterns extend into education, healthcare, engineering, science, entrepreneurship, and public services, the implications could be much larger than a modest productivity increase. People who currently lack access to elite training or expert support could gain tools that help them perform closer to the level of experienced practitioners.

That does not automatically produce equality, nor does it eliminate the need for human judgement. Access, governance, incentives, training quality, and concentration of power remain crucial questions. The same systems that spread expertise could also concentrate control if only a small number of organisations own the most capable models.

Even so, the novice-worker evidence matters because it provides a concrete example of a larger claim often made by advocates of AI abundance: that one of AI’s most important effects may be making useful intelligence less scarce. The strongest early workplace studies do not show everyone becoming an expert. They do suggest that many people can become competent much faster than before, and that may be one of the first visible steps toward a world where access to knowledge is far less limited by who already possesses it.

Amazon book picks

Further Reading

Books and field guides related to Why beginners may gain most from AI help. Use these as the next step if you want deeper reading beyond the article.

Book

Co-Intelligence

By Ethan Mollick

Directly relevant to how AI helps less experienced workers improve.

See on Amazon

Book

Working with AI

By Thomas H. Davenport, Steven M. Miller

Explores AI as an on-the-job assistant and coach.

See on Amazon

Book

Range

By David Epstein

Discusses how people acquire expertise across domains.

See on Amazon

Book

Peak

By Anders Ericsson

Explains how deliberate practice develops high-level skills.

See on Amazon

Browse more on Amazon:Co IntelligenceWorking with AIRange

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Shop location

UsingUSA USA USAUKCanadaAustraliaIreland

Example eBay listing

Retro Vintage Microphone Robot Desk Lamp, Vintage Microphone Robot Dimmer2367

SearcheBay.co.uk: desk robot

Browse similar oneBay.co.uk

Example eBay listing

Mini Robot Vacuum Cleaner – Cute Portable USB Rechargeable Desk Crumb & Dust -UK

SearcheBay.co.uk: desk robot

Browse similar oneBay.co.uk

Example eBay listing

Life-Size Robot Hand Phone Holder - Phone Stand - Various Colours - 3D Printed

SearcheBay.co.uk: desk robot

Browse similar oneBay.co.uk

Example eBay listing

Lavira Desk Robot Clock Freestanding Kids Novelty Home Decor Study Bedroom Gift

SearcheBay.co.uk: desk robot

Browse similar oneBay.co.uk

Browse more oneBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

1. Source: academic.oup.com
Link:https://academic.oup.com/qje/article/140/2/889/7990658

2. Source: nber.org
Link:https://www.nber.org/papers/w31161

3. Source: arxiv.org
Title: arXiv Generative AI at Work
Link:https://arxiv.org/abs/2304.11771

4. Source: gsb.stanford.edu
Link:https://www.gsb.stanford.edu/faculty-research/publications/generative-ai-work

5. Source: danielle.li
Title: GENERATIVE AI AT WORK˚
Link:https://danielle.li/assets/docs/GenerativeAIatWork.pdf

6. Source: siepr.stanford.edu
Link:https://siepr.stanford.edu/publications/working-paper/generative-ai-work

7. Source: mitsloan.mit.edu
Title: generative ai and worker productivity
Link:https://mitsloan.mit.edu/centers-initiatives/institute-work-and-employment-research/generative-ai-and-worker-productivity

8. Source: time.com
Title: How to Make AI Work for You, at Work
Link:https://time.com/6302984/ai-jobs-career/

9. Source: digitaleconomy.stanford.edu
Title: Digital Economy Lab Generative AI at Work
Link:https://digitaleconomy.stanford.edu/publication/generative-ai-at-work/

10. Source: arxiv.org
Title: arXiv AI and jobs. A review of theory, estimates, and evidence
Link:https://arxiv.org/abs/2509.15265

Published: September 18, 2025

11. Source: arxiv.org
Link:https://arxiv.org/abs/2502.08854

12. Source: nber.org
Link:https://www.nber.org/system/files/working_papers/w31161/w31161.pdf

13. Source: nber.org
Title: measuring productivity impact generative ai
Link:https://www.nber.org/digest/20236/measuring-productivity-impact-generative-ai

14. Source: nber.org
Link:https://www.nber.org/digest/202412/workplace-adoption-generative-ai

15. Source: nber.org
Title: Erik Brynjolfsson The Economics of Transformative AI. Book
Link:https://www.nber.org/people/erik_brynjolfsson

16. Source: brynjolfsson.com
Link:https://www.brynjolfsson.com/

17. Source: ideas.repec.org
Link:https://ideas.repec.org/p/ecl/stabus/4141.html

18. Source: economics.stanford.edu
Title: erik brynjolfsson
Link:https://economics.stanford.edu/people/erik-brynjolfsson

19. Source: gsb.stanford.edu
Title: generative ai work
Link:https://www.gsb.stanford.edu/faculty-research/working-papers/generative-ai-work

20. Source: arxiv.org
Link:https://arxiv.org/pdf/2304.11771

21. Source: Wikipedia
Title: Erik Brynjolfsson
Link:https://en.wikipedia.org/wiki/Erik_Brynjolfsson

22. Source: linkedin.com
Title: Erik Brynjolfsson
Link:https://www.linkedin.com/in/erikbrynjolfsson

Additional References

23. Source: bitter-sweet.ai
Link:https://www.bitter-sweet.ai/articles/generative-ai-at-work-brynjolfsson-2023/

24. Source: sternstrategy.com
Link:https://sternstrategy.com/speakers/erik-brynjolfsson/

25. Source: analysisgroup.com
Link:https://www.analysisgroup.com/people/affiliated-experts/erik-brynjolfsson/

26. Source: linkedin.com
Link:https://www.linkedin.com/posts/erikbrynjolfsson_great-to-see-generative-ai-at-work-my-activity-7297390624809357312-HX0R

27. Source: merriam-webster.com
Link:https://www.merriam-webster.com/dictionary/generative

28. Source: linkedin.com
Link:https://www.linkedin.com/posts/dejan-kovacevic1_w31161pdf-activity-7362723477138677760-MypQ

29. Source: itif.org
Link:https://itif.org/publications/2023/07/10/customer-support-agents-using-ai-gpt-tool-saw-nearly-14-percent-increase-in-productivity/

30. Source: linkedin.com
Link:https://www.linkedin.com/posts/joemckendrick_great-analysis-of-the-nber-working-paper-activity-7056978668945842177-6AWv

31. Source: laweconcenter.org
Title: ai productivity and labor markets a review of the empirical evidence
Link:https://laweconcenter.org/resources/ai-productivity-and-labor-markets-a-review-of-the-empirical-evidence/

32. Source: cfodive.com
Title: ai boosts productivity nber case study generative workforce
Link:https://www.cfodive.com/news/ai-boosts-productivity-nber-case-study-generative-workforce/649110/

Published: May 2023