TL;DR: AI workflows that cross device, identity and security boundaries create accountability gaps when approval becomes routine, Hexnode reports. Governance must connect ownership, escalation and recovery across tools before autonomous actions scale.
Editorial analysis by NHI Mgmt Group, based on content published by Hexnode: “HexCon26: Apu Pavithran and Ramy Nassar on Keeping Autonomous IT in Check”.
Key questions
Q: Why does autonomous IT create accountability risk even when humans approve the workflow?
A: Because approval does not transfer responsibility.
Q: How should security teams decide which AI decisions need human-in-the-loop review?
A: Start with impact, reversibility, and uncertainty.
Q: What breaks when human-in-the-loop approval becomes routine for AI agents?
A: The control breaks when approval stops being a real decision and becomes a reflex.
Practitioner guidance
- Define the accountable owner for every autonomous workflow Assign a named business and technical owner before the workflow is allowed to act, and make that owner responsible for scope, outcomes and rollback.
- Separate routine automation from high-risk approval gates Remove low-consequence actions from human approval queues and reserve sign-off for actions that change material state, access or customer impact.
- Validate autonomous actions against written operating procedures Require the system to compare its proposed action with documented SOPs before execution, especially for remediation, provisioning and recovery steps.
Bottom line: Autonomous IT is creating a governance problem because AI workflows now cross device, identity and security boundaries that were never designed to be managed as one system.
Glossary terms: Autonomous IT, Human-in-the-Loop (HITL), Governance Control Plane
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
Autonomous IT exposes a governance gap, not just an automation gap. The central problem is that organisations try to govern autonomous actions with tools built for human-paced approval and siloed administration. That approach breaks when devices, users, data and security controls are all touched in one workflow. The practical conclusion is that governance must be designed around outcomes, context and delegation, not around isolated tool approvals.
A question worth separating out:
Q: What should security teams do when autonomous workflows span multiple systems?
A: Treat the workflow as a governed control plane and require a single view of who approved it, what it accessed, what it changed and how it can be stopped. When multiple tools are involved, fragmented logging and approvals create blind spots, so the operating model must connect identity, device and security actions end to end.
👉 Read our full editorial: Autonomous IT governance now depends on accountable human oversight