美银美林发布软件行业趋势报告，提出AI 3.0即"界面层"时代：AI 1.0是模型，AI 2.0是智能体，AI 3.0是界面，用户通过自然语言聊天式界面表达意图并执行跨应用任务，应用不会消失但可能不再被直接访问。这引出价值归属问题：应用厂商凭深度嵌入、专有数据与工作流仍受益，但若用户日益通过外部AI界面交互，头部AI实验室可能沿栈下移、侵蚀应用层价值。报告认为AI部署还需要中间件层整合数据、治理与跨系统连接（IT、HR、CRM），并通过Copilot、Slack、ChatGPT、Claude等界面暴露能力；成本效率与统一架构整合能力将决定谁能获得增量AI支出。定价模式转向混合制：用户订阅加智能体消耗计费。
AI is changing not only what software can do, but how humans interact with it. If AI 1.0 was the model, AI 2.0 was the agent, then AI 3.0 is the interface, driving a meaningful shift in the relationship between users, applications and the software stack. Historically, enterprise software vendors owned every layer of the experience; agentic AI is beginning to disaggregate these layers, raising the question of where value accrues if the interface migrates to AI-powered assistants. Application vendors continue to benefit from entrenchment, proprietary data, workflows and business logic, but if users increasingly interact through external AI interfaces, leading AI labs could move down the stack and develop competing application layers. AI agents can understand intent, access context and execute work across applications, with the interface becoming the focal point. The deployment of AI is creating demand for a middleware layer that integrates data, governance and cross-system connections (I