Why Businesses Need Clearer Limits Before AI Agents Are Authorized to Act
An IBM survey of 2,000 C-level executives found only 11% feel fully prepared for AI agent deployment. PromptHalo founder Madhuri Chandoor argues the core issue is distinguishing capability from authority.
The Control Gap
A 2026 IBM study highlights concerns about AI readiness, visibility, and control. The survey revealed:
- Two-thirds of CIOs and CTOs are accountable for AI systems they do not fully control.
- 70% of teams are deploying technology faster than IT can track.
IBM presented these findings as an indication of a growing control gap as AI use expands across businesses.
Addressing the Control Gap with Contextual AI Security
Madhuri Chandoor, founder of PromptHalo, proposes an AI security and trust infrastructure that inspects why an action is being performed, not just what is being performed. Her approach aims to provide organizations with more contextual information about actions, considering user intent, assigned permissions, and surrounding circumstances.
The Refund-Splitting Example
Chandoor uses a refund scenario to illustrate the importance of context. An AI agent authorized to issue refunds up to $50 individually might be exploited by requesting ten $50 refunds instead of one $500 refund requiring review. Examining broader session context and behavior could help identify when escalation for human review is necessary, Chandoor suggests.
Behavioral Profiling for AI Agents
Chandoor draws parallels to financial fraud monitoring across transactions and accounts. She advocates developing behavioral profiles for autonomous agents, examining their actions over time, and documenting:
- What agents can access.
- Under what conditions.
- Potential downstream effects.