Beyond Autonomy: A Dynamic Tiered AgentRunner Framework for Governable and Resilient Enterprise AI Execution

TL;DR

This paper introduces a Dynamic Tiered AgentRunner framework for enterprise AI that enhances governability and resilience through risk-adaptive resource allocation, independent review, and failure recovery mechanisms.

Contribution

It presents a novel multi-mechanism framework combining risk-based tiering, agent separation, and resilience loops for safer enterprise AI deployment.

Findings

Achieves Pareto-optimal safety-efficiency trade-offs.

Ensures independent review and verification of AI proposals.

Provides a resilient execution loop with failure recovery.

Abstract

Current large language model agent frameworks prioritize autonomy but lack the governability mechanisms required for enterprise deployment. High-risk write operations proceed without independent review, complex tasks lack acceptance verification, and computational resources are allocated uniformly regardless of risk level. We propose the Dynamic Tiered AgentRunner, a controlled execution protocol distilled from a production-grade multi-tenant SaaS platform. The framework introduces three core mechanisms: (1) Risk-Adaptive Tiering that dynamically allocates computational resources and review intensity based on task risk profiles, achieving Pareto-optimal trade-offs between safety and efficiency; (2) Separation of Powers architecture where proposal, review, execution, and verification are performed by independent agents with physically isolated boundaries; and (3) Resilience-by-Design through a Verifier-Recovery closed loop that treats failure as a first-class system state. We formalize the tier selectio