Agentic migration

Expanded Definition

Agentic migration describes the organisational shift from using AI as a point solution toward embedding agents into end-to-end workflows where they can observe context, choose actions, and call tools or systems on behalf of people. The security significance is not the presence of AI itself, but the expansion of delegated authority across business processes, which makes identity, approval, monitoring, and rollback controls part of the design rather than an afterthought.

Definitions vary across vendors and practitioners because some teams use the phrase for simple workflow automation with an LLM in the loop, while others reserve it for systems where an OWASP Agentic AI Top 10 style threat model is needed due to autonomous tool use. NHI Management Group treats the term as a governance transition: the control problem moves from approving a model response to governing action execution, decision traceability, and permission scope across humans and software agents. That makes this concept closely aligned with NIST AI Risk Management Framework expectations around mapping, measuring, and managing AI risk. The most common misapplication is calling basic chatbot deployment agentic migration, which occurs when a system can generate text but cannot actually execute delegated actions in a controlled workflow.

Examples and Use Cases

Implementing agentic migration rigorously often introduces tighter authorisation and audit constraints, requiring organisations to weigh operational speed against the risk of uncontrolled action execution.

• A service desk agent creates tickets, queries asset data, and drafts remediation steps, but only executes changes after human approval on high-risk actions.
• A procurement workflow agent gathers vendor information, checks policy thresholds, and prepares purchase orders, while finance teams constrain its tool access to prevent unauthorised commitments.
• An SOC assistant correlates alerts, enriches incidents, and opens containment actions in limited playbooks, reflecting lessons highlighted in the Anthropic — first AI-orchestrated cyber espionage campaign report about the risks of abuse once agents can operate through tools.
• An identity operations agent reviews access requests, checks policy evidence, and recommends approvals, but must not self-approve privileged access without compensating controls.
• A customer operations agent updates records across CRM and billing systems, with step-up checks for actions that affect payment, privacy, or account status.

For security teams, the practical question is whether each delegated step is bounded, attributable, and reversible. That is why guidance from the OWASP Top 10 for Agentic Applications 2026 is useful when mapping tool abuse, prompt injection, and overbroad permissions to real workflow designs.

Why It Matters for Security Teams

Agentic migration matters because it changes the attack surface from a static application boundary to a living operational chain that includes prompts, tools, memory, data sources, and human override points. Once agents can act across systems, identity becomes central: each agent needs a defined principal, scoped privileges, and clear accountability for every decision and side effect. Without that, incident response becomes ambiguous, and basic questions such as who authorised the action, which policy allowed it, and how to stop recurrence become difficult to answer.

This is where NHI governance and agentic ai security intersect. If a software agent is treated like a user but managed like a script, organisations often end up with excessive standing access, weak segregation of duties, and poor traceability. Frameworks such as the MITRE ATLAS adversarial AI threat matrix and the CSA MAESTRO agentic AI threat modeling framework help teams reason about abuse paths, but they do not replace identity controls, approval logic, or logging discipline.

Organisations typically encounter the consequences only after an agent makes an unauthorised change, leaks data through a tool, or executes a workflow outside its intended scope, at which point agentic migration becomes operationally unavoidable to address.

Related resources from NHI Mgmt Group

• Agentic Supply Chain
• Who is responsible when agentic tooling makes a bad migration decision?
• What is Agentic AI and how does it differ from traditional generative AI?
• What NHI types do Agentic AI systems typically use?