AI and the Apprenticeship Gap: Rethinking Work for Experienced and Early-Career Talent in Professional Services

Geoffrey T. Sanzenbacher, PhD, is a professor of the practice of economics at Boston College and a research fellow at the Center for Retirement Research. Boston College placed 31st in U.S. News & World Report’s 2027 survey of national universities. Boston College shared the placement in a three-way tie with Tufts University and New York University.¹

The Center for Retirement Research has been called “the nation’s leading center on retirement studies” by The New York Times. Sanzenbacher studies job mobility, labor-force participation, retirement, and longer careers.

Q. MIT Sloan recently reported that frontier AI models can complete roughly 50% to 75% of tested text-based workplace tasks at a minimally sufficient level without edits, and that model failure rates on such tasks have been halving about every 2.2 to 2.8 years.² If AI capability is spreading broadly across many tasks rather than arriving through isolated breakthroughs, how should employers and workers age 55 and older distinguish between jobs that are expected to be redesigned and jobs at genuine risk of displacement over the next three to five years?

A. Sanzenbacher: The realistic distinction is between AI changing a set of work activities and AI removing the person from the workflow. A job is more likely to be redesigned when AI handles routine preparation but the worker remains accountable for defining the problem, determining if a response aligns with facts, recognizing exceptions, and crafting a recommendation.

In turn, a job is at greater risk when it consists mainly of predictable outputs from stable inputs and the employer uses AI only to cut headcount. Consider demand forecasting.

AI may quickly analyze sales, promotions, pricing, inventory, weather, and prior customer behavior. But an experienced employee may recognize that a new competitor has changed the market, that a supply interruption is being mistaken for weakened demand, or that a weather effect applies only to a particular product line.

AI can help test those possibilities, but it may not identify them unless a person knows enough to ask. This reality is why the durable value of experience is not tenure alone.

It is the ability to bring history, context, and skepticism to received LLM responses. Workers who can say, “This conclusion depends on an assumption we should test,” or “Our department may be measuring the wrong variables,” are doing work that is difficult to reduce to a standard prompt.

As such for employers, they need to be able to recruit, develop, and retain staff, who regardless of age, have the curiosity and capability to ask the next question that has not been asked yet.

As for older employees, my research does not show that AI is causing all older-worker job exits. It does show that, after ChatGPT’s launch, older workers in more AI-exposed jobs experienced relative increases in exits from employment and specifically in transitions to unemployment, rather than in exits from the labor force.³

Employers should take that result seriously and offer training, role redesign, and transition support instead of assuming every employee will adapt to AI in the same manner. For employees, regardless of age, the best response is not to compete with AI on speed.

It is to learn enough about the tool to evaluate it and enough about the business to challenge it. Spencer Stuart recently made a similar point about leadership in the AI era: the leaders most likely to thrive will learn quickly, ask better questions, and build trust.⁴

Q. In conversations with students I mentor through Harvard Business School’s and Boston College’s Mentor Match programs, and with Wharton undergraduates in my class, I hear both excitement about AI and real anxiety about entering the workforce. The Economist found that employment for 22- to 25-year-olds in AI-exposed occupations declined 13% relative to older workers between late 2022 and early 2025.⁵

If AI is taking over the routine assignments that have traditionally supported an apprenticeship model, through which new graduates develop judgment under the guidance of more experienced colleagues, how can employers redesign entry-level roles so that AI strengthens, rather than closes off, the path to professional judgment and advancement?

A. Sanzenbacher: When I worked for the Analysis Group, an economic consulting firm, there were two key assets that entry-level roles produced: intellectual capital which generated revenue for the near term and a pipeline of future higher-bill rate seasoned professionals. The intellectual capital was the research summary, spreadsheet, document review, first draft, or market analysis.

Today, AI can increasingly help create strong drafts of such content. As for the human capital, their expertise grows through doing that work and receiving feedback on key remits including how to assess evidence, form hypotheses, recognize weak assumptions, and craft compelling recommendations which clients value and are willing to pay premium fees.

If firms automate the intellectual capital but do not replace the learning process, they risk creating a thinner leadership pipeline. BCG’s 2026 survey of 70 C-suite leaders and senior executives found that half were already observing AI-related deskilling; more than 60% expected it to become a material threat within three to five years. The leaders identified judgment and decision making, problem understanding and framing, creative thinking, analysis and causal reasoning, and solution generation and evaluation as both important to long-term performance and vulnerable to deskilling.⁶

The solution is not to preserve manual work for its own sake. Rather, I believe it is to turn the remaining work into a more deliberate apprenticeship in questioning.

At this point, your readers might wonder how and why they can improve their questioning skills. I would argue that a legal framework could offer a useful model. Yale Law School teaches students to experiment with AI tools while developing a framework for critically evaluating them.

Yale Law’s Femi Cadmus frames the exercise through practical questions: Is the system secure? Where did the data come from? Is it reliable or “clean”? Has it been reviewed?⁷ This framework is another transferable method for employers to attack the apprenticeship knowledge gap.

Finally, don’t wait for employers to elevate your questioning skills and other experiences that can help support your career sustainability. Whether you are an accounting or philosophy major or somewhere in between, recognize that each person brings a unique set of experiences to the employment table.

It is up to you to take ownership of your career and tap into the various courseware that many leading organizations offer for free (https://pz.harvard.edu/resources/creative-question-starts) to society.⁸

Geoffrey T. Sanzenbacher, PhD, is a professor of the practice of economics at Boston College and a research fellow at the Center for Retirement Research. Boston College placed 31st in U.S. News & World Report’s 2027 survey of national universities. Boston College shared the placement in a three-way tie with Tufts University and New York University.⁸ The Center for Retirement Research has been called “the nation’s leading center on retirement studies” by The New York Times. Sanzenbacher studies job mobility, labor-force participation, retirement, and longer careers.

About the Author

Michael Wong is a part-time lecturer for the Wharton Communication Program at the University of Pennsylvania. He is an emeritus co-president and board member of the Harvard Business School Healthcare Alumni Association and a contributing writer for the MIT Sloan Career Development Office. His ideas have appeared in Harvard Business Review and MIT Sloan Management Review.

References

1. Dunn J. Boston College ranked 31st among best national universities. Boston College. Published September 2026. Accessed October 7, 2026. https://www.bc.edu/content/bc-web/sites/bc-news/articles/2026/fall/us-news-rankings.html

2. Murray S. How will AI automation hit—like a crashing wave or a rising tide? MIT Sloan Ideas Made to Matter. Published August 5, 2026. Accessed October 7, 2026. https://mitsloan.mit.edu/ideas-made-to-matter/how-will-ai-automation-hit-a-crashing-wave-or-a-rising-tide

3. Sanzenbacher GT. Are the careers of older workers being cut short by AI? Issue in Brief 26-13. Center for Retirement Research at Boston College; June 2026. Accessed October 7, 2026. https://crr.bc.edu/wp-content/uploads/2026/06/IB_26-13.pdf

4. Moioli F, et al. How leaders can thrive in the age of AI. Spencer Stuart. Published June 2025. Accessed October 7, 2026. https://www.spencerstuart.com/research-and-insight/how-leaders-can-thrive-in-the-age-of-ai

5. Don’t blame AI for your job woes. The Economist. Published November 6, 2025. Accessed October 7, 2026. https://www.economist.com/finance-and-economics/2025/11/06/dont-blame-ai-for-your-job-woes

6. Goel S, et al. When everyone uses AI, companies risk losing critical skills. Boston Consulting Group. Published June 17, 2026. Accessed October 7, 2026. https://www.bcg.com/publications/2026/when-everyone-uses-ai-companies-risk-critical-skills

7. Yale Law School. Yale Law School shapes the future of artificial intelligence. Published April 10, 2024. Accessed October 7, 2026. https://law.yale.edu/yls-today/news/yale-law-school-shapes-future-artificial-intelligence

8. Project Zero, Harvard Graduate School of Education. Creative question starts. Accessed October 7, 2026. https://pz.harvard.edu/resources/creative-question-starts