Goldman Sachs’ new hires will manage AI agents from day one

At Goldman Sachs, the first job title for a fresh hire may soon effectively be “manager.” The catch is that the direct reports are software.

Kevin Sneader, who heads the bank’s Asia Pacific business, said new finance hires will oversee AI agents from the very start of their careers. He made the remarks at the Milken Institute Asia Summit in Singapore on October 8, 2026.

What Goldman is actually changing

Instead of spending years grinding through grunt work before earning the right to delegate, new employees will be delegating from the outset. Their delegates are AI agents: tools that can take a task, break it into steps, and carry it out with a human checking the work.

Goldman’s push into agentic AI started in 2025 with Devin, a coding agent built by Cognition. The bank rolled it out across a technology division of about 12,000 engineers.

The scope has since widened. Goldman has deployed Anthropic’s Claude into core business functions, including trade accounting, compliance, and client onboarding.

The results so far include productivity gains of more than 20% in software engineering tasks since agentic AI was introduced.

Junior staff, senior skills

Chief Information Officer Marco Argenti has framed the change as a shift in what early-career employees need to learn. In his view, junior staff must get good at describing tasks, handing them off, and supervising AI systems — abilities that have traditionally been developed much later in a banking career.

This sits inside a broader internal program the bank calls OneGS 3.0. The initiative aims to redesign operational processes around teams that blend human workers and AI agents, rather than simply bolting new tools onto old workflows.

The middle manager problem

The source reporting points directly to a reduced need for current middle managers. If a junior hire can assign work to an agent, review the output, and move on, some of that coordination layer becomes harder to justify.

What this means for banks and the people who want to work at them

The research on Goldman’s strategy suggests the approach may attract a more tech-savvy class of employees and push financial firms toward skill-based roles rather than seniority-based ones.

There is a training risk, though. Junior years have traditionally doubled as an apprenticeship. If agents do the grinding, juniors need another way to build that intuition. Supervising an AI’s work is only useful if the supervisor knows what good work looks like.

Compliance and trade accounting are areas where mistakes draw regulatory attention. Handing more of that work to AI agents raises the stakes on the human oversight Goldman is counting on its newest employees to provide.

The things worth watching are concrete. One is whether Goldman’s productivity claims extend beyond engineering into the business functions where Claude is now deployed. Another is how the bank’s headcount mix shifts between junior, middle, and senior ranks over the next few hiring cycles.