Before artificial intelligence begins to replace jobs on Wall Street, it is first creating new positions.
Exclusive analysis provided to CNBC by Draup, a corporate recruitment data company, shows that banks including JPMorgan Chase, Citibank, and Capital One have seen a 49% increase in the number of job postings related to AI this year compared to 2025, reaching 139,819.
Draup indicates that the fastest-growing area is a set of skills related to AI agents. The company captures data from public recruitment information and platforms such as LinkedIn. For example, mentions of agent and orchestration – that is, the ability to design agents who can collaborate to complete tasks – have soared by 1,721% this year.
Draup CEO Vijay Swaminathan said in an interview, "This can be considered the hottest skill on Wall Street."
He said, "This is a huge opportunity; they need people who understand data, as well as AI, and who also know where to put it."
These job listings indicate that Wall Street banks are moving beyond chatbots and entering the next phase of their AI strategy, a phase that affects executives, employees, and shareholders alike. In order to fulfill their commitment to AI of improving productivity and automating repetitive tasks, banks are paving the way for a future where a large number of agents will undertake an increasing amount of work.
Previously, the recruitment focus of AI was mainly on engineers and data scientists, who were responsible for building models or adapting models to corporate data; however, the current trend has expanded to include those who are responsible for directly integrating AI into business operations.
Swaminathan indicates that when deploying AI within financial institutions, it is usually necessary to chain multiple specialized proxies together: for example, one proxy checks the raw data, another analyzes the files, and a third verifies regulatory compliance.
Employees involved in this process are usually referred to as Pre-Deployment Engineers ( forward-deployed engineers ), and they need to possess both technical capabilities and knowledge in specific business or functional areas, ranging from trading desks to back-office operations and human resources, says Swaminathan.
He said, "There is a lot of complexity within a company. Sometimes this complexity is obvious, but in many cases it is hidden. Even automating a simple process can take a long time."
For example, he said that if one were to establish a set of agents to automatically approve a company's leave applications, a network composed of various edge cases and specific exceptions would be formed.
Swaminathan indicates that agent and orchestration skills are particularly important for pre-deployment engineers, as their job is to figure out which proxies are needed, what each proxy does, and which technologies to use. He said that this also includes determining when manual supervision is required.
AI Technology Stack
Other popular skills related to the construction of AI, involving tools and technologies that help agents complete their work.
The mention of the framework LangGraph used for building multi-step workflows increased by 679%; LlamaIndex, which helps AI applications connect data, saw a growth of 291%; and retrieval-enhanced generation (RAG) – a technology that inputs company database information into AI models – experienced a 259% increase in mentions.
However, in addition to technical capabilities, companies are also placing increasing emphasis on so-called soft skills.
He said, "Our analysis shows that soft skills such as problem-solving, creativity, the ability to ask sharp questions, and maintaining perseverance while gaining a deeper understanding of processes are once again being valued."
AI Other areas of growth in this field also include those responsible for establishing safeguards for emerging systems.
According to the data from Draup, the frequency of mentioning "responsible AI" in job listings has soared by 657% this year; the mentions of AI governance and risk management have increased by 394% and 359% respectively. Security teams are also paying attention to prevent systemic vulnerabilities caused by third-party tools or external models being integrated.
In the data of Draup, governance-related skills are now mentioned over 16,000 times, which is nearly twice as many as the approximately 8,400 mentions related to model training, deployment, and operation.
He said, "We are very concerned about ensuring that the third parties we use in these products do not get out of control from a cybersecurity perspective."
According to the data from Draup, positions related to generative AI and proxy services generally have higher salaries than other technical positions in the financial industry. The median basic annual salary for managers in generative AI is approximately $190,000.
Despite higher salaries, filling these professional positions remains challenging, says Swaminathan.
To fill this gap, large banks are vigorously promoting internal retraining programs to train existing developers and domain experts, he said. Such efforts will also have a chain reaction: JPMorgan Chase's CEO Jamie Dimon once mentioned that as AI takes on more responsibilities, the company will undertake a "major redeployment plan."
Swaminathan said, "I believe that if we give priority to these soft skills with the right technical capabilities, people will adapt and learn. This is an exciting era for the right talents."
