麦肯锡-数据领导者扩展人工智能的操作指南（英）

September 2024McKinsey TechnologyA data leader’s operating guide to scaling gen AIDeploying generative AI in the enterprise requires a data-centric road map. Leaders can use a well-defined operating model to successfully scale the technology.This article is a collaborative effort by Alex Singla, Asin Tavakoli, Holger Harreis, Kayvaun Rowshankish, and Klemens Hjartar, with Gaspard Fouilland and Olivier Fournier, representing views from McKinsey Technology and QuantumBlack, AI by McKinsey.After almost two years of infatuation with generative AI (gen AI), companies are moving past the honeymoon phase1 to embrace the work that matters most: creating value from this tantalizing technology. Expectations are high. A recent McKinsey Global Survey found that 65 percent of companies across sizes, geographies, and industries now use gen AI regularly, twice as many as last year.2 Investment in gen AI continues to rise amid the belief that early gains seen by high performers are a harbinger of cost decreases and profits to come. But most companies have not yet seen significant impact from gen AI.To keep up with the competitive pace of innovation, data executives at most organizations have drafted gen AI strategies. Not all companies have moved past the pilot stage, but most have made steps to integrate AI into their tech stacks at some level. Yet a technical integration model is only part of what is necessary to generate lasting value from gen AI. Companies must also create gen AI operating models to ensure their technology implementations deliver measurable business results.An operating model is a familiar structure in most large organizations. A company’s operating model is a plan that outlines how people, processes, and technology will be deployed to provide value to customers and stakeholders. It can encompass financial structures, partnerships, and product road maps to meet the company’s long-term goals. When applied specifically to gen AI, an operating model includes every decision—from staffing and organizational structures to technology development and compliance—that guides how gen AI is used and measured throughout a company.A well-defined gen AI operating model can help leaders successfully and securely scale gen AI across their organizations. Data is the backbone of a successful gen AI deployment, so chief data officers (CDOs) often lead the charge to create these models—bringing technology, people, and processes together to transform gen AI’s potential into real impact. Yet when creating gen AI operating models, data leaders commonly fall into two traps:—Tech for tech: This approach involves allocating significant resources toward gen AI without a clear business purpose, leading to solutions disconnected from real-world impact. This can result in overspending on gen AI tools that are rarely used in daily workflows and create little business value.—Trial and error: This approach entails experimenting with disparate gen AI projects, but not doing s