Using Calculated Autonomy for AI Agents

Session Summary

In this session from MoEngage’s Agentic Marketing Tour, Angela Vega (Expedia), Ankur Kothari (GEICO), and Alesia Miller (Honeywell Technologies) discuss how to replace guesswork with a structured, risk-based framework for AI agent autonomy. Rather than relying on subjective guidelines, the panel explains how measuring the true cost of being wrong dictates how much operational independence an agent should receive. They examine how to separate high-risk legal and transactional constraints from flexible creative tasks, why campaigns reaching millions require mandatory human sign-off regardless of AI confidence scores, and how automated guardrails enable scalable, compliant personalization across complex audience segments.

Top Takeaways

• Adopting AI Through Narrow Use Cases: Mitigate operational and regulatory risk by testing AI agents in targeted, bounded workflows before integrating them into larger production environments.

• Automating Compliance and Quality Review: Utilizing specialized AI review agents to pre-audit messaging against regulatory standards significantly reduces legal bottlenecks while preserving compliance.

• Maintaining Human-in-the-Loop Oversight: In high-consequence industries, retaining mandatory human review on all customer-facing touchpoints protects brand integrity and prevents compliance failures.

• Centralizing Governance with Dedicated Councils: Establishing cross-functional oversight councils ensures every AI initiative aligns with company standards and risk tolerance before receiving approval to scale.

• Evolving Beyond Static Triggers: Shifting from rigid, rule-based automation to dynamic agentic workflows requires modernizing operational mindsets and attribution models alongside technology.