The shift to review capacity
Google reports that 75% of all new code is AI-generated and engineer-approved, a figure that rose from 50% last fall and 30% in April 2025. One complex migration finished 6x faster than a year ago. Snap expects to save $500 million annually after cutting 1,000 employees, as 65% of its new code is now AI-generated. This news caused Snap stock to rise 7%. These shifts move the headcount math from engineering capacity to review capacity. When AI handles the majority of new code, the discount factor on senior reviewers increases. Goldman Sachs restricted advanced tools like Anthropic’s Claude Code and Microsoft’s GitHub to engineers only, citing productivity and risk concerns. This decision came after the bank’s CIO mentioned "vibe coded" projects by traders and bankers. You should realize that as companies redirect budgets toward compute and automation, the technical requirements for remaining staff increase.
| Company | AI Code % | Workforce Impact |
|---|---|---|
| 75% | 75% of new code is AI-generated | |
| Snap | 65% | 16% workforce reduction |
| Meta | N/A | 8,000 jobs cut in AI restructuring |
The business request cycle drops from three days to 30 minutes by removing handoffs, data extraction, and translation between professional languages. LLMs can translate a business question into data queries, retrieve knowledge from documents, and orchestrate existing tools to return results with the context needed for a final decision. The tech industry is moving money from payroll to AI infrastructure. Meta cut 8,000 jobs in an AI restructuring wave, with 10% of staff in middle management and non-engineering roles being removed to fund a $115-135B AI capex year. Microsoft offered voluntary buyouts to 8,750 US employees at the senior director level and below. UKG laid off 950 employees in an "AI-first" restructuring. The 2026 tech layoff wave has seen more than 100,000 jobs cut worldwide by early May, including 81,700 in the first quarter and 20,000 more in the first six weeks of the second quarter. Cisco is cutting thousands of roles after record revenue as part of a $1 billion restructuring plan.
Growing demand for agent orchestration
The demand for skills involving AI agents exploded this year. References to agent orchestration jumped 1,721% according to analysis from Draup. This skill involves designing agents that work together on specific tasks to handle complex processes that once required multiple human handoffs between different professional languages and data models. The demand for LangGraph grew 679%, while mentions of LlamaIndex rose 291% and retrieval-augmented generation rose 259%. Wall Street banks are moving beyond chatbots to a strategy of deploying armies of agents. This requires forward-deployed engineers who combine technical abilities with domain knowledge of trading desks or back-office operations.
Posts for AI-related roles at banks including JPMorgan Chase, Citigroup and Capital One rose 49% this year to 139,819 listings. References to "responsible AI" rose 657%, while mentions of AI governance grew 394% and AI risk management grew 359%. New niches like AI hiring-system auditors are forming due to EU AI Act enforcement rules. The complexity of an enterprise makes automation difficult, such as creating a team of agents to automate employee vacation requests. This requires determining which agents are needed, what each one does and which technology to use. It also involves deciding when human overseers need to be involved. In the cryptocurrency market, an estimated 65% of all trading volume in 2026 involves some form of automation. Can these automated agents replace the nuance of a human trader during a market crash?
The new talent profile
The profile of a successful candidate has changed. Recruiting in the quant space became PhD-centric following the 2009 financial crisis. Candidates in the R&D spectrum with a CS or ML PhD from a renowned professor can command total compensation exceeding seven figures. Newly-minted computer science graduates can field offers of $300,000. Median base salaries for generative AI managers sit at $190,000. Some funds now prioritize prompt engineering or "vibe coding" effectiveness. Moelis is training a model to pinpoint attributes like warmth and curiosity to move past the "like-me" bias that traditionally limited the talent pool.
| Role Type | Compensation Detail |
|---|---|
| AI Research PhD | Seven figure total compensation |
| CS Graduate | $300,000 base salary |
| GenAI Manager | $190,000 median base |
The hunt for talent is intense. Competition centers on a small pool of fewer than 1,000 AI research scientists who build advanced large language models. OpenAI’s average stock-based compensation reached $1.5 million among its 4,000 employees in 2025. At Series D startups, employee stock grants can range from $2 million to $4 million. Jensen Huang noted at Stanford GSB that most people will lose their job to somebody who uses AI rather than to AI itself. One hedge fund uses AI as the primary trader, where researchers feed the reasoning engine to make better decisions. This reality forces firms to invest in internal reskilling programs to train existing developers and domain experts.




