AI Job Openings at Wall Street Banks Rise 49% for 2026

Job postings related to artificial intelligence at banks such as JPMorgan Chase, Citigroup, and Capital One totaled 139,819 in 2026, a 49% increase over 2025. The survey was conducted by Draup, a corporate hiring data company, and was exclusively published by CNBC this Friday (2).

The aggregate number matters less than the composition. Mentions of "agent orchestration," the ability to design agents that work together on the same task, grew by 1,721% over the year. "This is likely the most sought-after skill on Wall Street," said Vijay Swaminathan, CEO of Draup, to CNBC.

Governance Surpasses Model Building

The most revealing data appears in another column. According to Draup, governance competencies appear over 16,000 times in the job postings analyzed from the banks, compared to about 8,400 references to training, deploying, and operating models. The ratio is nearly two to one. Posts mentioning "responsible AI" increased by 657%, those mentioning AI governance by 394%, and risk management by 359%.

The reading is that banks are hiring more people to supervise agents than to build them. It makes sense for those tracking the design of the systems described by CNBC: chains of specialized agents that check raw data, review documents, and ensure outputs meet regulatory requirements. Each link in this chain needs a human responsible for responding to the auditor.

For Swaminathan, the demand goes beyond model engineers: "It's a huge opportunity; they need people who understand data and people who understand AI and where to place it." The fastest-growing profile, according to the survey, is that of the forward deployed engineer, a professional who integrates agents directly into trading desks, compliance areas, and back-office operations.

The Regulatory Vacuum That Banks Fill Alone

Interestingly, the governance movement does not respond to a formal requirement. On April 17, the Federal Reserve, the OCC, and the FDIC published a review of the interagency guidance on model risk management, known as SR 26-2. The text states that generative and agent-based AI models "are new and evolving rapidly," and, therefore, "are not within the scope of this guidance." The agencies promised to address the topic in future guidance.

Thus, banks are setting up control structures before knowing what rules they will need to comply with. Those hiring now define the internal standard that will likely serve as a reference when the regulator writes the rule, giving larger banks an agenda advantage over medium-sized institutions that are waiting for the final text.

What the Numbers Don’t Tell

There are clear limitations to the data. Draup measures job postings collected from public portals and LinkedIn, not actual hires. The same position may be posted more than once, and the 1,721% increase is based on a base that the survey does not detail. Additionally, the survey does not address the other side of the equation: how many operations, customer service, and document verification positions were not opened during the same period.

This is the point that separates the optimistic reading from the skeptical. A 49% increase in AI job postings coexists with workforce reduction programs at the same banks: Citigroup, one of the banks cited by Draup, has been executing a planned cut of 20,000 positions since 2024. Both can be true at the same time, and the metric of open positions captures only half of the equation.

Where Demand Lands

Nominal hiring occurs in New York, but much of the execution does not. JPMorgan, Citi, and Goldman Sachs maintain large technology centers in India, in cities like Bengaluru, Hyderabad, and Pune. It is there that many governance and agent operation roles are likely to settle, as they are continuous functions intensive in review, the type of work these centers already perform for risk and compliance.

In Poland, Citi operates a service and technology center in Warsaw that serves the bank's global network. A migration of back-office tasks to agent chains directly impacts this type of unit: some verification work disappears, and some transforms into agent supervision, requiring a different professional profile.

For the CIO, the question shifts from how many AI engineers the bank has to who signs off on a decision made by an agent, and whether that person is already on the payroll or still an open position.