The paper argues that embedding an intelligent agent into an existing application creates a fundamental coordination problem: users can change goals or shared objects while the agent continues executing. This can lead to situations where individual actions are correct but the overall effect no longer matches the user's intent. To address this, the authors propose (AIS) as a pattern that combines a conventional application core with a built-in agent, and Intent-Level Interaction Abstraction (IIA) as the user-facing semantics for inspecting and controlling delegated tasks. They introduce interaction contracts that constrain the relationship between task-level interaction and application behavior through task bindings, authority, control transitions, and outcome evidence. Continuous assurance maintains these claims as dependencies change. The paper presents this as a framework and research agenda for making agent integration a maintainable engineering discipline.
Embedded software agents need interaction contracts and continuous assurance
The authors formalize how applications can keep delegated agent work aligned with changing user goals, permissions, shared state, and evidence.
Top University
Shengcheng Yu · Chunrong Fang · Zhenyu Chen
Technical University of Munich · State Key Laboratory for Novel Software Technology · Nanjing University
Research Digest··2 min read
In one line
Dependable agent integration requires explicit contracts linking task-level interaction to application behavior and continuous assurance.
Before you read this
Two ideas this paper builds onagent-integrated softwareSoftware whose behavior is driven by an embedded AI agent, enabling dynamic adaptation, autonomous task execution, and contract-based integration.3 papers →runtime monitoringAudits an AI agent's live actions, internal states, or effects against constraints or policies to catch errors and attacks.107 papers →
Read it as
Yu, Fang, and Chen propose Agent-Integrated Software, a pattern combining a conventional application core, direct user interaction, and a built-in agent.
The authors define (AIS) and an Intent-Level Interaction Abstraction (IIA), through which users can inspect and control delegated tasks.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.