Device-Native Autonomous Agents for Privacy-Preserving Negotiations

TL;DR

This work introduces a device-native autonomous negotiation framework that preserves privacy by executing entirely on user hardware and securing interactions with zero-knowledge proofs and cryptographic audit trails. It combines six architectural innovations—selective state transfer, explainable memory, world-model distillation, a privacy-preserving negotiation protocol, model-aware offloading, and simulation-critic safety—within an 8-step agentic workflow. Empirical results in insurance and B2B contexts show high success rates (87–90%), up to 2.4× latency improvements over cloud baselines, and substantial privacy preservation (as low as 14 bits leaked). The approach demonstrates strong user trust benefits from explainability and provides a foundation for trustworthy, privacy-centric autonomous agents in regulated financial domains.

Abstract

Automated negotiations in insurance and business-to-business (B2B) commerce encounter substantial challenges. Current systems force a trade-off between convenience and privacy by routing sensitive financial data through centralized servers, increasing security risks, and diminishing user trust. This study introduces a device-native autonomous Artificial Intelligence (AI) agent system for privacy-preserving negotiations. The proposed system operates exclusively on user hardware, enabling real-time bargaining while maintaining sensitive constraints locally. It integrates zero-knowledge proofs to ensure privacy and employs distilled world models to support advanced on-device reasoning. The architecture incorporates six technical components within an agentic AI workflow. Agents autonomously plan negotiation strategies, conduct secure multi-party bargaining, and generate cryptographic audit trails without exposing user data to external servers. The system is evaluated in insurance and B2B procurement scenarios across diverse device configurations. Results show an average success rate of 87%, a 2.4x latency improvement over cloud baselines, and strong privacy preservation through zero-knowledge proofs. User studies show 27% higher trust scores when decision trails are available. These findings establish a foundation for trustworthy autonomous agents in privacy-sensitive financial domains.

Figures (6)

• Figure 1: Agentic AI workflow architecture showing 8 sequential stages from Goal Initiation to Outcome Evaluation, with Tool Hub and dual-memory (STM/LTM) components.
• Figure 2: Component integration flow through six technical innovations during negotiation execution.
• Figure 3: Success rate vs. scenario complexity for insurance and B2B. Performance degrades gracefully with complexity.
• Figure 4: Latency breakdown by component across scenarios. Ins represents Insurance, B2B represents business-to-business, S/M/C represent Simple/Medium/Complex.
• Figure 5: Radar chart comparison across 5 metrics (higher values indicate better performance; latency is inverted). The proposed system achieves balanced performance across all dimensions.