Architecting Human-AI Cocreation for Technical Services -- Interaction Modes and Contingency Factors

TL;DR

This paper develops a six-mode taxonomy of human-AI interaction in technical services, linking interaction modes to contingency factors to guide the design of safer, more effective collaborative systems.

Contribution

It introduces a comprehensive framework connecting interaction modes with contingency factors, aiding practitioners in designing appropriate human-AI collaboration levels.

Findings

Proposes a six-mode taxonomy of human-AI interaction.

Links interaction modes to task complexity, risk, and system reliability.

Provides a systematic method for selecting human oversight levels.

Abstract

Agentic AI systems, powered by Large Language Models (LLMs), offer transformative potential for value co-creation in technical services. However, persistent challenges like hallucinations and operational brittleness limit their autonomous use, creating a critical need for robust frameworks to guide human-AI collaboration. Drawing on established Human-AI teaming research and analogies from fields like autonomous driving, this paper develops a structured taxonomy of human-agent interaction. Based on case study research within technical support platforms, we propose a six-mode taxonomy that organizes collaboration across a spectrum of AI autonomy. This spectrum is anchored by the Human-Out-of-the-Loop (HOOTL) model for full automation and the Human-Augmented Model (HAM) for passive AI assistance. Between these poles, the framework specifies four distinct intermediate structures. These include the Human-in-Command (HIC) model, where AI proposals re-quire mandatory human approval, and the Human-in-the-Process (HITP) model for structured work-flows with deterministic human tasks. The taxonomy further delineates the Human-in-the-Loop (HITL) model, which facilitates agent-initiated escalation upon uncertainty, and the Human-on-the-Loop (HOTL) model, which enables discretionary human oversight of an autonomous AI. The primary contribution of this work is a comprehensive framework that connects this taxonomy to key contingency factors -- such as task complexity, operational risk, and system reliability -- and their corresponding conceptual architectures. By providing a systematic method for selecting and designing an appropriate level of human oversight, our framework offers practitioners a crucial tool to navigate the trade-offs between automation and control, thereby fostering the development of safer, more effective, and context-aware technical service systems.