Service providers compose services in service chains that require deep integration of core operational information systems across organizations. Additionally, advanced analytics inform data-driven decision-making in corresponding AI-enabled business processes in today’s complex environments. However, individual partner engagements with service consumers and providers often entail individually negotiated, highly customized Service Level Agreements (SLAs) comprising engagement-specific metrics that semantically differ from general KPIs utilized on a broader operational (i.e., cross-client) level. Furthermore, the number of unique SLAs to be managed increases with the size of such service chains. The resulting complexity pushes large organizations to employ dedicated SLA management systems, but such ‘siloed’ approaches make it difficult to leverage insights from SLA evaluations and predictions for decision-making in core business processes, and vice versa. Consequently, simultaneous optimization for both global operational process efficiency and engagement-specific SLA compliance is hampered. To address these shortcomings, we propose our vision of supplying online, AI-supported SLA analytics to data-driven, intelligent core workflows of the enterprise and discuss current research challenges arising from this vision. Exemplified by two scenarios derived from real use cases in industry and public administration, we demonstrate the need for improved semantic alignment of heavily customized SLAs with AI-enabled operational systems. Moreover, we discuss specific challenges of prescriptive SLA analytics under multi-engagement SLA awareness and how the dual role of AI in such scenarios demands bidirectional data exchange between operational processes and SLA management. Finally, we discuss the implications of federating AI-supported SLA analytics across organizations.

服务提供者在服务链中组合服务这需要跨组织深度集成核心运营信息系统。此外，在当今复杂的环境中，高级分析为相应的支持 AI 的业务流程中的数据驱动决策提供信息。然而，单个合作伙伴与服务消费者和提供者的约定通常需要单独协商的、高度定制的服务水平协议 (SLA)，其中包含特定于约定的指标，这些指标在语义上不同于在更广泛的运营（即跨客户）级别上使用的一般 KPI。此外，要管理的唯一 SLA 的数量会随着此类服务链的规模而增加。由此产生的复杂性促使大型组织采用专用的 SLA 管理系统，但是，这种“孤立”的方法使得很难利用 SLA 评估和预测中的洞察力进行核心业务流程的决策，反之亦然。所以，同时优化全球运营流程效率和特定于参与的 SLA 合规性受到阻碍。为了解决这些缺点，我们提出了为企业的数据驱动、智能核心工作流提供在线、人工智能支持的 SLA 分析的愿景，并讨论了这一愿景带来的当前研究挑战。以从工业和公共管理中的实际用例派生的两个场景为例，我们证明了需要改进高度定制的 SLA 与支持 AI 的操作系统的语义对齐。此外，我们讨论了在多参与 SLA 意识下规范性 SLA 分析的具体挑战，以及 AI 在此类场景中的双重角色如何要求操作流程和 SLA 管理之间的双向数据交换。最后，我们讨论了跨组织联合 AI 支持的 SLA 分析的影响。