Core Insights - Deutsche Bank's report emphasizes that companies are still in the early stages of developing their AI transformation roadmaps after attending the 2025 New York AI Leaders Summit [1] Group 1: Investment Return and Data Readiness - There is a lack of consensus on measuring return on investment (ROI), with business leaders defining their own key metrics [2] - Data readiness remains critical, with management's understanding of data and its storage locations being fundamental issues [2][3] - Only 10%-20% of total time is reportedly spent on training models, indicating that the quality of models heavily depends on input data [2] Group 2: AI Implementation and Governance - Approximately 80% of clients are still in the phase of optimizing existing business processes, while 20% are more willing to experiment [2] - Regulatory and governance policies are seen as barriers to the speed of AI adoption across enterprises [3] - Many leaders believe that maintaining human involvement in agent processes is crucial for reasonableness checks [2][3] Group 3: Preferences and Strategies - There is an increasing preference for small language models (SLMs) over large language models (LLMs) due to better control and efficiency [3] - Multi-cloud strategies appear to be the preferred approach, with leaders favoring a "best of breed" method [3] - Low-risk and repetitive workloads are identified as the first areas to leverage AI, with significant opportunities for value creation in backend functions [3] Group 4: Survey Insights - 73% of participants believe their organizations are making uneven progress in AI application journeys, with only 18% of systems in production and 9% in early pilot stages [4] - 70% prioritize balancing AI innovation with security, while the rest distribute focus among rapid deployment, risk management, and regulatory compliance [4] - The biggest barrier to creating seamless AI-driven customer experiences is legacy system integration (56%), followed by unclear ROI (33%) and data silos (11%) [4]
从2025纽约AI领袖峰会看企业AI落地:多云策略与小模型成主流选择