TL;DR: Agentic AI systems shift governance from prompt review to identity, delegation and tool access control, according to Black Kite. Protocols like MCP and A2A make standing credentials and weak authorisation the main risk boundary for runtime autonomy.
Editorial analysis by NHI Mgmt Group, based on content published by Black Kite: “Understanding Agentic AI & Protocols:”.
Key questions
Q: Why do agentic AI workflows create new IAM risk compared with traditional automation?
A: Traditional automation usually follows fixed rules and predictable paths, so its access model is easier to review.
Q: What breaks when organisations treat agent identities like service accounts?
A: What breaks is accountability.
Q: How should security teams govern AI agents that can access enterprise systems?
A: Security teams should govern AI agents as non-human identities with explicit ownership, scoped privileges, and continuous monitoring.
Practitioner guidance
- Map every agent authority path Document which tools, datasets, and external actions each agent can reach, then identify where authority is delegated, inherited, or reissued.
- Separate protocol access from model access Treat MCP or A2A connectivity as a control boundary and require explicit approval for each tool or peer connection the agent can invoke.
- Inventory agent credentials as NHIs Track tokens, API keys, and service credentials used by agent workflows with the same lifecycle controls you apply to workload identities.
Bottom line: Agentic AI changes the security problem from prompt control to authority control, because the system can take actions rather than only generate text.
Glossary terms: Agentic AI, Model Context Protocol, Delegated Authority Model, Identity Sprawl
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
Agentic AI governance starts with identity, not prompts. The core problem is no longer whether a model can generate a useful answer. It is whether a software actor can obtain, carry, and spend authority in a way that is bounded enough for enterprise governance. Prompt-centric controls miss the real question because the control plane has moved to delegation, tool access, and runtime decision-making. Practitioners should treat the agent itself as the thing being governed.
A few things that frame the scale:
- The agentic AI market is already worth $5 billion and is projected to reach $50 billion by 2030.
A question worth separating out:
Q: What is the difference between an AI agent and a normal automated workflow for security teams?
A: A normal workflow follows predetermined logic, while an AI agent may interpret instructions and choose actions at runtime. That difference matters because the security boundary shifts from a fixed process to a trust relationship that can be manipulated through content, prompts, or delegated context.
👉 Read our full editorial: Agentic AI protocols change how identity and access are governed