Agentic Service Value Model
How the ITIL 4 Service Value System adapts when agents become first-class participants in every value chain activity
ITIL 4’s Service Value System (SVS) provides a useful organizing model for how IT service management components work together to create value. The SVS describes five components: guiding principles, governance, the service value chain, practices, and continual improvement. Together, they describe how demand becomes value through organized service management activity.
That model was designed assuming humans performed the activities within each value chain step. It remains valid as an organizing framework. But when agents become active participants in those activities, each component of the SVS must be understood differently. This section defines how.
The Original Service Value System
The ITIL 4 SVS centers on a Service Value Chain with six activities: Plan, Improve, Engage, Design and Transition, Obtain and Build, and Deliver and Support. These activities interact with 34 management practices to convert demand and opportunity into products and services that deliver value.
In the human-centric model, the value chain is executed by IT professionals who make decisions, coordinate across teams, and perform the work of each activity. Automation assists at specific, defined points. Humans remain in the loop at every decision.
How Agentic Participation Changes the Model
The Agentic Service Value Model extends the SVS rather than replacing it, defining a new category of participant, the AI agent, and clarifying how agent participation changes each component.
In the agentic model, agents are present in every value chain activity. They are not concentrated in Deliver and Support, though that is where current deployment maturity is highest. The full value of agentic ITSM is realized when agents participate across the entire chain.
Guiding Principles in the Agentic Context
The seven principles in Section 1 of this framework define the agentic extension of ITIL 4’s guiding principles. The original ITIL 4 principles (focus on value, start where you are, progress iteratively, collaborate and promote visibility, think and work holistically, keep it simple and practical, optimize and automate) remain applicable and are complemented by the agentic-specific principles that address the unique challenges of autonomous participants.
The critical addition is the distinction between optimization and autonomy. ITIL 4’s “optimize and automate” principle addresses efficiency. The agentic context requires a further question: not just “how do we automate this step?” but “what should agents decide, and what should humans decide?”
Governance in the Agentic Context
ITIL 4 defines governance as the means by which an organization is directed and controlled. In the human-centric model, governance applies to human actors, defining their authority, accountability, and the standards they must meet.
In the agentic model, governance must apply equally to agents. Agents make decisions. Agents take actions. Agents have access to sensitive systems and data. Governance that addresses human actors but not agent actors is governance with a gap large enough to produce serious incidents.
The full governance model is defined in Section 7. The key governance additions specific to the agentic context are:
Agent identity: Every agent must have a defined identity, separate from any human identity, within every system it accesses. Agent actions must be attributable to a specific agent identity in audit logs, not to a shared service account.
Autonomy tiers: Agents should be classified by their autonomy level, from assisted (agents recommend, humans decide and act) through collaborative (agents act, humans review) to autonomous (agents act within defined parameters without human review). Different tiers require different governance controls.
Action boundaries: For every agent, there must be a current, versioned document that defines what the agent may do autonomously, what requires human approval before execution, and what the agent must never do.
Override authority: Every agent must have a functioning kill switch and a clear escalation path to human review. These must be tested in production, not just documented.
The Service Value Chain in the Agentic Context
Each value chain activity changes materially when agents participate.
Plan
In the human-centric model, planning is a periodic, calendar-driven activity performed by senior practitioners. Capacity plans are produced quarterly. Risk assessments are conducted annually. Demand forecasts are built from historical data with manual analysis.
In the agentic model, planning becomes continuous. Capacity agents model resource utilization across every infrastructure layer in real time and produce demand forecasts that incorporate business calendar data, project pipelines, and external signals. Risk assessment agents continuously evaluate the change calendar, infrastructure health, and known vulnerability data. Planning is no longer a meeting; it is an ongoing, data-driven process that surfaces findings to humans when decisions are required.
Improve
In the human-centric model, improvement is driven by periodic reviews, service improvement programs, and individual analyst insight. Identifying what to improve depends on humans noticing patterns in data they have the capacity to review.
In the agentic model, improvement detection is automated. Agents continuously analyze incident patterns, resolution effectiveness, SLA performance trends, knowledge base usage, and CMDB drift. They identify improvement opportunities and surface them to human process owners with supporting evidence. The problem statement for improvement programs is delivered by agents; the decision to invest and the design of the improvement remain human responsibilities.
Engage
In the human-centric model, engagement is through service desk channels where humans interpret requests, qualify them, and route them to the appropriate team.
In the agentic model, engagement is conversational and immediate. Users interact through whatever channel they already use, natural language in collaboration tools, email, voice, mobile. Agents interpret intent, qualify the request against policy and entitlement data, and either resolve it directly or initiate the appropriate fulfillment workflow. Engagement agents operate 24 hours a day, 365 days a year, without queue degradation at high volume.
Design and Transition
In the human-centric model, change analysis relies on practitioner knowledge of CI dependencies, and transition validation relies on human execution of pre-defined checklists.
In the agentic model, impact analysis is automated using real-time CMDB graph data. Change agents identify affected services, users, and downstream dependencies with a completeness that human analysts cannot match in the time available. Deployment validation agents monitor production health during and after transitions, execute rollback automatically when thresholds are breached, and update CMDB records as changes complete.
Obtain and Build
In the human-centric model, provisioning and fulfillment require human execution across multiple systems. Access provisioning means an analyst logging into an identity provider, executing configuration steps, and confirming completion.
In the agentic model, fulfillment is orchestrated by agents across connected systems. A service request for software access triggers an agent that validates entitlement, executes provisioning in the identity system, deploys software through endpoint management tooling, updates the CMDB, and notifies the user on completion, autonomously and within minutes rather than days.
Deliver and Support
This is where agentic deployment is currently most mature, and the gap between current practice and the agentic model is most visible. The domain pages in Section 4 document this gap for each of the twelve ITSM practice domains.
Practices in the Agentic Context
ITIL 4 defines 34 management practices organized into general management practices, service management practices, and technical management practices. In the agentic model, practices must define agent participation alongside human participation.
This means practice documentation requires two additions. First, for each practice activity, the definition of which aspects are appropriate for agent execution and which require human judgment. Second, for each agent-executed activity, the data dependencies, integration requirements, and governance controls that must be in place before agent execution is authorized.
The twelve practice domains covered in Section 4 of this framework provide this documentation for the twelve ITSM practices with the highest impact from agentic transformation.
Continual Improvement in the Agentic Context
ITIL 4’s Continual Improvement model provides a cycle for identifying, planning, and executing improvements to services and practices. In the agentic context, this cycle must be applied to agents themselves.
Agents are not static. Their performance changes as the data they reason over evolves, as the environment they operate in changes, and as edge cases accumulate that were not anticipated during initial design. Organizations that do not continuously monitor and improve their agents will experience performance drift, and that drift will compound because agents operate at machine speed.
The Continual Improvement cycle applied to agents looks like this: Define agent performance baselines and targets. Monitor agent performance continuously against those targets. Identify the root causes of performance gaps, whether data quality issues, model drift, integration failures, or incorrect action boundaries. Design and implement improvements. Validate improvement effectiveness. Repeat.
The Metrics and Measurement section in Section 6 provides the measurement framework for this continual improvement cycle.
Summary
The Agentic Service Value Model is the ITIL 4 SVS with agents recognized as first-class participants in every component. It changes who, and what, executes the activities within the SVS, not what the SVS itself describes. Organizations that understand and design for this change will build ITSM programs that compound in capability over time. Those that add agents without rethinking the model will add cost without proportionate value.
References
Axelos. (2019). ITIL Foundation: ITIL 4 Edition. TSO.
Dumas, M., Milani, F., & Chapela-Campa, D. (2026). Agentic Business Process Management Systems. arXiv preprint arXiv.18833.
Maes, S. H. (2026). Agentic Smart ITIL, And The Disruption Of The Market Of Conventional Enterprise Applications. Stephane H. Maes’ Blog on WordPress / Multi-Agent Research Notes.