Agent Orchestration Skills Surge 1,721% as Wall Street Banks Rewire AI Hiring

Before artificial intelligence takes Wall Street jobs, it is creating them — 139,819 of them this year, to be exact. That is the number of AI-related job postings at banks including JPMorgan Chase, Citigroup and Capital One, a 49% increase compared with 2025, according to an analysis by enterprise hiring data firm Draup provided exclusively to CNBC.

The single fastest-growing skill cluster in those postings has nothing to do with building models. It is agent orchestration — the ability to design multiple AI agents that work in concert on a task. References to agent orchestration in bank job listings jumped 1,721% this year, Draup found.

"This is arguably the hottest skill on Wall Street," Draup CEO Vijay Swaminathan said in an interview. "It's a massive opportunity. They need people who understand data and people who understand AI and where to put it."

Draup culls its data from public job posts and platforms including LinkedIn. The numbers describe a specific shift in how banks deploy AI — and why the shift matters well beyond bank HR departments.

From chatbots to agent armies

The job listings show that Wall Street banks are moving beyond chatbots to the next phase of their AI strategy. That phase carries implications for executives, employees and shareholders. To deliver on AI's promised productivity gains and automation of repetitive tasks, banks are pressing toward a future in which armies of agents handle an increasing share of the labor.

An earlier wave of AI hiring was dominated by engineers and data scientists who built models or adapted them to corporate data. The current boom has expanded to include people responsible for embedding AI directly into business lines.

Deploying AI inside a financial institution often requires stringing together multiple specialized agents: one to inspect raw data, another to analyze a document and a third to check regulatory compliance.

The workers who do this — often called forward-deployed engineers — need a mix of technical ability and domain knowledge of a specific business or function, from trading desks to back-office operations and human resources, according to Swaminathan.

Consider a deceptively simple example. Creating a team of agents to automate approval of employee vacation requests creates a web of edge cases and specific exemptions, Swaminathan said.

"There is a lot of complexity in an enterprise," he said. "Sometimes these complexities are visible, but many times they are hidden. It takes a long time even to automate a simple process."

Agent orchestration sits at the center of the forward-deployed engineer's job description. These workers figure out which agents are needed, what each one does and which technology to use. The role also involves deciding when human overseers need to step in, Swaminathan said.

The agent tech stack

The Draup data shows banks are hiring for concrete tools, not just vague AI fluency.

References to LangGraph, a framework for building multistep workflows, jumped 679%. Mentions of LlamaIndex, which helps connect AI applications to data, rose 291%. References to retrieval-augmented generation, or RAG — a technique for feeding AI models information from company databases — climbed 259%.

These are the plumbing skills of the agent era. LangGraph structures how agents hand off tasks. LlamaIndex connects them to the data they need. RAG grounds their outputs in a bank's own records. Together, they form the toolkit that turns a demo into a system a regulated institution can run.

Soft skills are back in demand

Beyond technical abilities, banks are putting growing emphasis on soft skills — and Draup's data backs that up with a pointed observation about what hiring managers now ask for.

"Our analysis shows that there is a renewed focus on soft skills like problem solving, creativity, ability to ask tough questions, being assertive [when it comes to] deeper understanding of the processes," Swaminathan said.

The logic is straightforward. An agent system is only as good as the process map a human builds for it. If the human cannot ask tough questions about how a workflow actually runs — including the informal workarounds nobody documents — the agents will automate the wrong thing.

Guardrails and governance

Another growth area involves creating guardrails around the emerging systems, including demand for risk and control infrastructure.

References in job postings tied to "responsible AI" surged 657% this year, according to Draup. Postings mentioning AI governance jumped 394%. Those mentioning AI risk management climbed 359%. Security teams are also focused on preventing third-party tools or external model connections from creating systemic vulnerabilities.

The scale of governance demand is notable. Governance-related skills now account for more than 16,000 references in the Draup data — nearly twice the roughly 8,400 references tied to training, deploying and running models.

For banks, this ratio reflects regulatory reality. Financial institutions operate under strict compliance regimes, and every AI system that touches customer data, trading decisions or employee records inherits those obligations. The hiring data suggests banks now worry as much about controlling AI as deploying it.

"There is a lot of focus on making sure that the third parties that we are using in these products are not going to rogue from a cybersecurity standpoint," Swaminathan said.

The pay gap — and the talent gap

Roles tied to generative AI and agents typically pay more than tech roles elsewhere in finance. Generative AI managers earn a median base salary of about $190,000, according to Draup.

Even at that price, banks struggle to fill these specialized roles, Swaminathan said. The combination of technical depth, domain knowledge and process judgment is rare.

To bridge the gap, major banks are leaning heavily into internal reskilling programs to train existing developers and domain experts, he said. That approach has a second-order effect: it prepares current employees for the displacement AI is expected to cause.

JPMorgan CEO Jamie Dimon has spoken of "huge redeployment plans" as AI takes over more work. The reskilling programs Draup describes are the practical machinery behind such plans.

"I think the more we prioritize those soft skills with the right amount of technical skills, people will adapt and learn," Swaminathan said. "It's a very exciting time for the right talent."

Why it matters

The Draup numbers quantify a turn in the AI labor market. The first phase of enterprise AI rewarded people who could build models. This phase rewards people who can wire models into the messy, regulated, exception-ridden processes of a global bank — and control the result.

The 1,721% surge in agent orchestration references, the near-doubling of governance skills over model-operations skills, and the $190,000 median salary for generative AI managers all point in the same direction: banks now treat AI agents as production infrastructure, not experiments. The workers who can orchestrate, govern and oversee those agents will set the pace of how quickly Wall Street's redeployment — Dimon's "huge" one — actually happens.