Generative artificial intelligence is projected to affect tens of millions of workers by automating many entry-level roles while simultaneously lowering the barriers to more technically demanding positions. A report from the Burning Glass Institute and Harvard Business School’s Project on Managing the Future of Work finds that generative artificial intelligence could impact 50 million US workers by automating millions of entry-level jobs, yet it may also enable candidates without post-secondary credentials or extensive experience to perform higher-skill tasks through artificial intelligence tools. This shift challenges the 150-year-old pyramid model of organizations, in which large cohorts of junior workers gain experience and rise into leadership, and suggests an “artificial intelligence at the bottom” structure that still must find ways to give people the oversight and judgment-building experience that artificial intelligence cannot replicate.
The report concludes that artificial intelligence-based automation could make nearly 18 million entry-level jobs in the US obsolete, accounting for about 12% of the total workforce, in occupations such as legal associates, marketing specialists, and project managers. At the same time, about 29 million “mastery roles” will become accessible to more workers as artificial intelligence tools lower technical skill requirements in positions like network administrators, data warehousing specialists, and loan interviewers. Joseph Fuller argues that firms cannot simply “crimp” the bottom of the talent ladder, because without fresh cohorts of junior analysts and specialists, organizations will lack people with enough experience to supervise agentic artificial intelligence systems, spot hallucinations, and set policies. Artificial intelligence can make highly accurate, rules-based decisions where rich historical data is available, but organizations will still need seasoned humans to make final calls on capital investments and risk, drawing on contextual judgment and knowledge of strategy and culture.
Companies are experimenting with artificial intelligence mainly as an efficiency layer on top of existing processes, which often delivers only marginal gains and fuels disappointment, especially when poor-quality, disorganized internal data undermines outcomes. Fuller argues that organizations need to redesign workflows from the ground up to exploit a general purpose technology rather than treat artificial intelligence as a bolt-on. At the same time, restricted access to major artificial intelligence platforms and limited training mean many employees rely on older free versions at home and use them like search engines, resulting in underutilization. Research cited in the report shows that people are 50% more likely to use artificial intelligence at home than at work, highlighting a corporate training gap. To keep up with rapid technological change, employers will have to pivot from a traditional “school first, job later” model to work-based learning, with intense, bursty, and frequent training and more cooperative education and apprenticeship-style experiences that build both technical fluency and social skills long before promotion to leadership roles.
As artificial intelligence permeates core business functions, large companies such as JP Morgan Chase, Coca-Cola, and Cisco are already rebuilding processes like marketing and loan origination on an artificial intelligence foundation. This deep integration is shifting hiring criteria toward soft skills and adaptability: employers will prioritize social skills, the capacity to learn quickly, and grit, with firms like Accenture explicitly valuing determination and resilience. Fuller expects that within a few years, candidates will routinely face questions about how many agentic tools they have built, what went wrong, how they inherited and adapted tools created by others, and which models they combine in sequence for different applications, supported by portfolios of multimedia work. Because this is described as the first technology that “hits harder the further you go up the earning spectrum,” traditional middle-skill job displacement narratives no longer fit, and existing policies for displaced workers are characterized as an “expensive, dismal failure.” Policymakers are urged to focus on incentives that encourage citizens and employers to ease the transition rather than attempt to impose top-down solutions that markets can easily circumvent.
