Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GAIA: A General Agency Interaction Architecture for LLM-Human B2B Negotiation & Screening

Siming Zhao, Qi Li

TL;DR

GAIA tackles safe, auditable LLM‑human delegation in high‑stakes B2B screening and negotiation by enforcing a governance‑first protocol. It introduces a three‑role model (Principal, Delegate, Counterparty) with an optional Critic, and three mechanisms—information‑gated progression via Task‑Completeness Index, dual feedback channels, and formal authorization boundaries with escalation—to prevent unauthorized commitments and ensure information sufficiency. The framework is formalized as a typed state machine with four safety invariants and a preflight safety layer, and is validated through a hybrid blueprint combining automated metrics and human judgment across procurement, real estate, and staffing scenarios. A staffing instantiation demonstrates practical gating, non‑binding language, and escalation, and the evaluation plan provides a path toward enterprise‑ready deployment. GAIA thus offers a reproducible specification for safe, efficient, and auditable AI delegation in enterprise workflows.

Abstract

Organizations are increasingly exploring delegation of screening and negotiation tasks to AI systems, yet deployment in high-stakes B2B settings is constrained by governance: preventing unauthorized commitments, ensuring sufficient information before bargaining, and maintaining effective human oversight and auditability. Prior work on large language model negotiation largely emphasizes autonomous bargaining between agents and omits practical needs such as staged information gathering, explicit authorization boundaries, and systematic feedback integration. We propose GAIA, a governance-first framework for LLM-human agency in B2B negotiation and screening. GAIA defines three essential roles - Principal (human), Delegate (LLM agent), and Counterparty - with an optional Critic to enhance performance, and organizes interactions through three mechanisms: information-gated progression that separates screening from negotiation; dual feedback integration that combines AI critique with lightweight human corrections; and authorization boundaries with explicit escalation paths. Our contributions are fourfold: (1) a formal governance framework with three coordinated mechanisms and four safety invariants for delegation with bounded authorization; (2) information-gated progression via task-completeness tracking (TCI) and explicit state transitions that separate screening from commitment; (3) dual feedback integration that blends Critic suggestions with human oversight through parallel learning channels; and (4) a hybrid validation blueprint that combines automated protocol metrics with human judgment of outcomes and safety. By bridging theory and practice, GAIA offers a reproducible specification for safe, efficient, and accountable AI delegation that can be instantiated across procurement, real estate, and staffing workflows.

GAIA: A General Agency Interaction Architecture for LLM-Human B2B Negotiation & Screening

TL;DR

GAIA tackles safe, auditable LLM‑human delegation in high‑stakes B2B screening and negotiation by enforcing a governance‑first protocol. It introduces a three‑role model (Principal, Delegate, Counterparty) with an optional Critic, and three mechanisms—information‑gated progression via Task‑Completeness Index, dual feedback channels, and formal authorization boundaries with escalation—to prevent unauthorized commitments and ensure information sufficiency. The framework is formalized as a typed state machine with four safety invariants and a preflight safety layer, and is validated through a hybrid blueprint combining automated metrics and human judgment across procurement, real estate, and staffing scenarios. A staffing instantiation demonstrates practical gating, non‑binding language, and escalation, and the evaluation plan provides a path toward enterprise‑ready deployment. GAIA thus offers a reproducible specification for safe, efficient, and auditable AI delegation in enterprise workflows.

Abstract

Organizations are increasingly exploring delegation of screening and negotiation tasks to AI systems, yet deployment in high-stakes B2B settings is constrained by governance: preventing unauthorized commitments, ensuring sufficient information before bargaining, and maintaining effective human oversight and auditability. Prior work on large language model negotiation largely emphasizes autonomous bargaining between agents and omits practical needs such as staged information gathering, explicit authorization boundaries, and systematic feedback integration. We propose GAIA, a governance-first framework for LLM-human agency in B2B negotiation and screening. GAIA defines three essential roles - Principal (human), Delegate (LLM agent), and Counterparty - with an optional Critic to enhance performance, and organizes interactions through three mechanisms: information-gated progression that separates screening from negotiation; dual feedback integration that combines AI critique with lightweight human corrections; and authorization boundaries with explicit escalation paths. Our contributions are fourfold: (1) a formal governance framework with three coordinated mechanisms and four safety invariants for delegation with bounded authorization; (2) information-gated progression via task-completeness tracking (TCI) and explicit state transitions that separate screening from commitment; (3) dual feedback integration that blends Critic suggestions with human oversight through parallel learning channels; and (4) a hybrid validation blueprint that combines automated protocol metrics with human judgment of outcomes and safety. By bridging theory and practice, GAIA offers a reproducible specification for safe, efficient, and accountable AI delegation that can be instantiated across procurement, real estate, and staffing workflows.

Paper Structure

This paper contains 34 sections, 2 equations, 3 figures, 1 table, 6 algorithms.

Table of Contents

  1. Introduction
  2. Motivating Scenario
  3. The Agency Challenge
  4. Gaps in Existing Approaches
  5. Our Approach: GAIA Overview
  6. Key Contributions
  7. Paper Organization
  8. Related Work
  9. LLM Negotiation and Bargaining
  10. Multi-Agent Architectures
  11. Human-in-the-Loop Agent Systems
  12. Safety and Authorization in Agentic AI
  13. Positioning and Gap Summary
  14. The GAIA Framework
  15. Reader's Map
  16. ...and 19 more sections

Figures (3)

  • Figure 1: GAIA role architecture showing five roles (Principal, Delegate, Counterparty, Critic, Moderator) and their information flows. The Delegate operates within authorization boundaries, tracks TCI for information completeness, and escalates to the Principal on boundary violations or safety events.
  • Figure 2: GAIA state machine with information-gated transitions. States are color-coded: blue (active), green (terminal success), orange (terminal neutral), red (escalation). The TCI-based gate controls the SCREEN $\rightarrow$ NEGOTIATE transition, while escalation paths (dashed red) enable human intervention at any point.
  • Figure 3: Dual learning channels coordination with precedence ordering. Three input streams (AI Critic, human micro-feedback, safety signals) are merged with explicit precedence: human overrides take highest priority, followed by safety signals, clarity tactics, and persuasion tactics. The merged context respects budget constraints.

Theorems & Definitions (6)

  • definition 1: Agency Task
  • definition 2: Screening vs Negotiation
  • definition 3: Authorization Boundary
  • definition 4: Task-Completeness Index, TCI
  • definition 5: Information Gain, IG
  • definition 6: Screening Efficiency