From Workflow Automation to Capability Closure: A Formal Framework for Safe and Revenue-Aware Customer Service AI

TL;DR

This paper introduces a formal framework to ensure safety and revenue-awareness in customer service AI systems that dynamically compose multiple specialized agents, addressing emergent safety risks.

Contribution

It presents a novel formal framework for verifying safety in complex, multi-agent customer service AI systems, bridging a critical safety gap in current platforms.

Findings

Identifies safety risks in multi-agent AI compositions.

Proposes a formal verification approach for safe agent interactions.

Addresses the emergent safety gap in customer service automation.

Abstract

Customer service automation is undergoing a structural transformation. The dominant paradigm is shifting from scripted chatbots and single-agent responders toward networks of specialised AI agents that compose capabilities dynamically across billing, service provision, payments, and fulfilment. This shift introduces a safety gap that no current platform has closed: two agents individually verified as safe can, when combined, reach a forbidden goal through an emergent conjunctive dependency that neither possesses alone.