The Mindset Shift Needed to Thrive Alongside AI | Marco Argenti, Goldman Sachs | RAISE Summit 2026

Organizational Context and Scale

• Goldman Sachs employs approximately 45,000 staff, with 12,000 full-time engineers and thousands of contractors, meaning one in three employees is an engineer.
• The firm operates in a highly regulated financial sector but leverages its fully digital nature to pursue aggressive AI adoption, comparing its constraints to Formula One racing where regulation drives performance.
• AI implementation touches every business aspect, focusing on developer communities, process reinvention, and client enablement.

Productivity Gains and Metrics

• The CIO reports a consistent 20% net productivity increase across the engineering organization.
• Unlike previous years where engineering teams requested additional headcount, recent reviews show teams requesting more resources to handle increased scope, indicating "self-funding" of projects.
• Traditional productivity metrics (lines of code, merge requests) are deemed unreliable for AI-enhanced workflows; the firm prioritizes observing step changes in output and early project completion.
• Code generation currently accounts for 30–40% of a developer's workload, with significant remaining opportunities in testing, deployment, and DevOps observability.

Shift to Prototype-Driven Development

• The development lifecycle has shifted from spec-driven to prototype-driven, significantly reducing the back-and-forth between business units and engineers.
• Business professionals (e.g., traders, wealth managers) now use "vibe-coding" tools (including Devin, Cognition, Cloud Code, and Codex) to build functional prototypes in days rather than weeks.
• Engineers no longer receive napkin sketches but rather precise, functional reproductions of business concepts, allowing them to focus on "hardening" and scaling production-ready code.
• This approach increases the fidelity of the final product by ensuring the prototype matches the client's vision before engineering resources are applied.

Leadership and Governance Strategy

• CEO David Solomon actively uses AI tools and models, establishing a "top-down" signal that complements "bottom-up" innovation.
• The firm mitigates the risk of uncontrolled innovation by providing a safe platform with strict guardrails against hallucinations, data leakage, and unauthorized access.
• An internal network of "AI Champions" and "AI Bar Raisers" fosters peer-to-peer adoption, creating a culture of comfort around experimentation and questions.
• Compensation is tied to overall outcomes rather than specific AI usage metrics, as AI adoption is viewed as an enabler of better business performance.

Safety and Operational Protocols

• Goldman Sachs treats model output as "junior human" input, subjecting all AI-generated artifacts to a rigorous Software Development Lifecycle (SDLC) with CI/CD pipelines, code reviews, and gates.
• Technical safeguards focus on "grounding" LLMs to prevent fabrication, ensuring models only reference actual data or utilize tools when the output is reachable.
• The firm maintains a robust testing environment to catch unexplainable behavior before models reach production.

Impact on Roles and Future Workflows

• The role of junior bankers is shifting from performing mechanical analysis (80–90% of time) to client engagement and apprenticeship, facilitated by the "Banker Co-Pilot" tool.
• This rebirth of apprenticeship accelerates learning by allowing junior staff to analyze data and create presentations at unprecedented speeds.
• CIOs and executives utilize the internal "GSA Assistant" (wired to 40,000 users and 2 million monthly prompts) for rapid research and client context.
• While the CIO delegates all research and information gathering to agents, he retains exclusive authority over final decision-making, asserting that human judgment remains superior to AI for strategic decisions.