Agentic AI Is Coming for the Entry-Level Economy

AI agents that can do whole tasks are quietly erasing the first rung of the career ladder in knowledge and creative work.

What to Know

• AI could expose up to 300 million jobs globally to automation over the next decade.
• In the U.S., AI could automate roughly 25% of all work hours, displacing 6–7% of workers over about 10 years.
• AI has already shaved around 16,000 jobs a month from U.S. payroll growth even as productivity rises.
• Companies can now rent “agents as a service” for coding, finance, customer service, and design work.
• Entry-level workers in their 20s and 30s in knowledge and content roles are most exposed to early job losses.

Artificial intelligence used to be a tool that helped professionals write emails or brainstorm ideas. In 2026, it is increasingly something closer to a digital worker. Goldman Sachs chief information officer Marco Argenti says models are now “entities or agents that can perform tasks on your behalf,” able to browse, reason across documents, and take multistep actions. That shift from chatbot to “agentic AI” is where the pressure on entry-level roles begins.

Goldman Sachs Research estimates that AI could automate tasks representing roughly 25% of U.S. work hours, displacing 6–7% of workers over a decade in its base case. If adoption stays gradual, unemployment only nudges higher. If firms roll out agents aggressively, the risk is a “jobless recovery” in which output and profits climb while early‑career hiring stalls.

How Agentic AI Hits the First Rung

Agentic AI is landing first where work breaks into repeatable digital tasks, which is exactly how many entry-level knowledge jobs are structured.

Joseph Briggs, Co-Head of Global Economics, Goldman Sachs Research

Economist Joseph Briggs notes that AI’s initial impact has been concentrated in tech and digital media, where employment shares have slipped below long-term trends. The exposed roles are the classic first rungs into the middle class: junior coders, content writers, graphic designers, call-center agents, and junior consultants. Their work consists of research, drafting, and customer interactions that can be mapped into tasks for an AI agent.

Image generated by DALL-E, Bar chart showing AI exposure by age

Goldman’s follow-up analysis finds that AI has already removed about 16,000 jobs a month from U.S. payroll growth while measured productivity rises. For younger workers trying to land a first knowledge-sector job, it means fewer postings and a higher bar for showing value beyond what an agent can do.

Research from Stanford’s Digital Economy Lab points in the same direction. In occupations highly exposed to AI, entry-level hiring declined 13% relative to less-exposed jobs within firms after the spread of large language models. When a task can be handed entirely to an agent, the entry-level human doing that task is often the one who disappears first.

From “Software as a Service” To “Agent as a Service”

The business model around AI is shifting from renting tools to renting completed work, and that shift points straight at junior roles.

Marco Argenti, Chief Information Officer, Goldman Sachs

According to Goldman Sachs Chief Information Officer Marco Argenti:

“Now a year later, we look at models as essentially entities or agents that can perform tasks on your behalf.”

In a Fox Business interview, Argenti also described the shift toward “agent as a service,” where companies can deploy AI agents for coding, finance, customer service, and design. That means a finance team may use an agent to read filings and draft memos, while a design team may use one to generate concepts and format them for review.

Those are precisely the tasks historically given to interns, analysts, and junior associates. When a manager can ask an AI system to prepare a first-draft deck and get a passable product in minutes, the pressure to hire an entry-level generalist weakens. The risk is not that every junior role disappears overnight, but that each new headcount request has to clear a higher hurdle than a rented agent.

Argenti emphasizes that adaptability is now a core job skill. Workers who thrive will be those willing to rethink their day-to-day habits and learn how to supervise and improve agents rather than compete with them task by task.

Where AI Erases Jobs And Where It Creates Them

AI is compressing demand for some entry-level roles while pulling new demand into infrastructure, specialized care, and high-touch services.

The Goldman Sachs team is explicit that AI does not only destroy jobs. They highlight several areas of growth that offset part of the automation shock. The build-out of energy and data-center infrastructure is already lifting employment: construction jobs exposed to data centers have risen by about 216,000 since 2022, and the U.S. power sector may need roughly 500,000 net new workers by 2030 to support AI-driven electricity demand. These are skilled trades like electricians, HVAC technicians, and lineworkers.

AI also increases specialization inside high-skilled sectors such as healthcare. As technology takes over generic paperwork, new niche occupations emerge around specific procedures and patient needs. At the same time, as productivity lifts incomes, service roles expand in areas like pet care, tutoring, and athletic coaching, where Goldman notes about 1 million workers now employed in jobs that barely existed 30 years ago.

The problem is timing and fit. The entry-level content designer displaced by an AI agent is not automatically qualified to wire a data center or manage a high-end HVAC system. Without deliberate reskilling and clear pathways into these growth fields, the risk is a cohort of young workers stuck between disappearing junior roles and new opportunities they are not yet trained to fill.

Wrap Up

For American households, the rise of agentic AI means the economy can look healthy on the surface while early-career workers face a much rougher climb. GDP and corporate earnings can grow, productivity can rise, and stock indexes can rally even as entry-level postings in knowledge and creative sectors thin out. That is what economists mean by a jobless recovery: growth without broadly shared job gains.

Whether that scenario becomes reality is not predetermined. Policy can push training toward trades and infrastructure, firms can design junior roles that teach workers how to manage agents, and individuals can focus on skills that complement what an AI agent does well. The sooner that shift happens, the more likely it is that agentic AI becomes a ladder, not a wall.