2024-09-23-麦肯锡-数据领导者扩展人工智能的操作指南（英）_9页_1mb

McKinsey Guide to Scaling Generative AI: A Data-Centric Approach

• Generative AI (gen AI) adoption is rising, with companies moving beyond hype to focus on value creation through technology, people, and processes. Data leaders must avoid common pitfalls, such as "tech for tech" solutions (allocating resources without clear business purpose) and uncoordinated trials.

• A well-defined operating model is essential for scaling gen AI successfully and securely. This model, centered on data, includes decision-making frameworks for deployment, measurement, and integration with business goals.

• Key elements include designing gen AI around components for scalability, prioritizing data management (especially unstructured data), and establishing risk governance to handle AI-specific risks like misinformation or hallucinations.

• Team structures can be extended from existing IT/data teams or built as distinct entities, with a central CoE for expertise and governance. Gradually decentralizing development can improve agility.

• Development approaches vary from centralized (quick capability build) to federated or decentralized models, depending on industry regulations and business needs, ensuring alignment with organizational goals.

• Risk planning involves identifying, classifying, and mitigating potential threats through a tiered approach, with ongoing monitoring and compliance updates to address evolving regulations.

• To compete in the fast AI landscape, data leaders should create an agile operating model, focusing on data organization, deployment strategies, and continuous improvement to deliver measurable business results.

For detailed implementation, McKinsey emphasizes a step-by-step guide to building gen AI operating roadmaps, integrating data governance and risk management as foundational components.