Comparative Analysis of Approaches to Embedding AI Functionality through In-Application Interfaces and Autonomous Agent Orchestrators in CRM Systems

Engineering notes

Key topics: autonomous driving. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Customer relationship management platforms now embed artificial intelligence in two structurally different ways, and the choice between them shapes cost, governance, and the pace of organizational change. The first way places assistive models within the interfaces teams already use, so that scoring, summarization, and drafting occur at the point where work is performed. The second way introduces a separate orchestration tier in which semi-autonomous agents pursue goals across several systems while users remain in familiar screens. This review examines the two approaches through a seven-layer reference architecture that places each in a specific tier and clarifies why the boundary between them lies between the model and orchestration tiers. The analysis treats both as design options with measurable trade-offs across change management, expressiveness for multi-step work, architectural coupling, and the surface available for monitoring. Evidence from enterprise practice and the research literature suggests that embedded assistance is more successful during first-generation adoption in mature organizations because it leverages existing processes, whereas a dedicated agent tier yields stronger long-term outcomes for second-wave programs that target genuine cross-system automation. Supporting comparisons across data access, integration style, the trust model, and migration strategy show that the central decision sits inside a wider system of architectural commitments that the framework makes explicit for practicing architects.

Links and sources

• PDF from original source

Comments