OpenAI Research Details Enterprise Workflow Shifts as Workers Adopt Cross-Role AI Tasks

On September 16, 2026, OpenAI's Economic Research Team released the second installment of its 'Work at the Frontier' series, titled 'How workers are unlocking new ways of working'. Analyzing more than 1.5 million work-related ChatGPT interactions recorded between April and July 2026, the research establishes that knowledge workers are actively engaging in 'task crossover'—leveraging generative AI to execute responsibilities traditionally reserved for external departments or specialized roles. The study reveals that workers prompt models differently when operating outside their primary domain, submitting shorter prompts and relying on AI as borrowed expertise rather than requesting instructional tutorials. Notably, routine communication and promotional copywriting showed high adoption persistence, whereas highly regulated duties like financial analysis and legal research saw lower permanent crossover. For enterprise architects, engineering leaders, and operations executives, these findings indicate that the primary organizational friction point in generative AI adoption has shifted. The conversation is no longer about isolated seat-based productivity gains; it is about the quiet erosion of siloed enterprise functions. When line-of-business operators execute preliminary cross-functional workflows autonomously, departmental handoffs decline, but enterprise risk shifts downstream. AI-assisted task sprawl directly impacts governance, data security, and compliance posture when employees generate marketing copy or draft customer communications without traditional cross-departmental gatekeeping. This shift fits into the broader enterprise trend toward outcome-driven AI and agent-assisted workflows. Over the past two years, organizations deployed generative AI tools primarily for internal task acceleration within specific job descriptions. However, as foundation models improve in domain reasoning, individual knowledge workers increasingly function as multi-disciplinary generalists. Rather than replacing specialized teams, AI serves as an orchestration layer enabling non-specialists to handle first-pass drafts and operational coordination across departmental boundaries. In practice, technical leaders and DevOps/platform engineering teams should align their internal AI governance frameworks to account for cross-functional usage patterns. IT and security teams must implement fine-grained data loss prevention (DLP) and role-based access controls across enterprise AI gateways, ensuring that workers executing out-of-domain tasks do not inadvertently route sensitive data through unsecured channels. Furthermore, enterprise software procurement and HR workforce planning must re-evaluate job architecture, shifting from hyper-rigid functional silos toward collaborative, AI-augmented workflow boundaries that validate AI-assisted outputs before they enter core production systems.