Paper Type

Complete

Paper Number

PACIS2026-1925

Description

The Internet of Things generates vast streams of temporal data, yet translating these streams into actionable knowledge remains a persistent design challenge for Information Systems research. Generative AI, particularly Large Language Models, introduces new possibilities for interpreting, mediating, and augmenting such data. However, the emerging literature at this intersection is fragmented across technical domains and lacks systematic consolidation from a sociotechnical design perspective. This paper presents a structured review of 35 recent studies combining Generative AI with IoT and adjacent time series contexts. A two-stage analytical strategy pairs AI-assisted topic modeling with interpretive concept coding to identify four recurring themes: temporal intelligence through hybrid architectures, agentic mediation between users and sensor infrastructures, explainability demands in anomaly detection, and generative augmentation beyond forecasting. The synthesis reveals strong model-centric experimentation alongside underdeveloped organizational, individual, and societal design dimensions. These imbalances are consolidated into a research agenda for future artifact-oriented IS inquiry.

Comments

10-IoT

Share

COinS
 
Jul 5th, 12:00 AM

When Models Meet Sensors – A Structured Review and IS Research Agenda for Generative AI-Enabled IoT Artifacts

The Internet of Things generates vast streams of temporal data, yet translating these streams into actionable knowledge remains a persistent design challenge for Information Systems research. Generative AI, particularly Large Language Models, introduces new possibilities for interpreting, mediating, and augmenting such data. However, the emerging literature at this intersection is fragmented across technical domains and lacks systematic consolidation from a sociotechnical design perspective. This paper presents a structured review of 35 recent studies combining Generative AI with IoT and adjacent time series contexts. A two-stage analytical strategy pairs AI-assisted topic modeling with interpretive concept coding to identify four recurring themes: temporal intelligence through hybrid architectures, agentic mediation between users and sensor infrastructures, explainability demands in anomaly detection, and generative augmentation beyond forecasting. The synthesis reveals strong model-centric experimentation alongside underdeveloped organizational, individual, and societal design dimensions. These imbalances are consolidated into a research agenda for future artifact-oriented IS inquiry.