What is agentic ESM? Agentic enterprise service management explained

Agentic ESM (agentic enterprise service management) is enterprise service management in which AI agents, not people, do most of the work of resolving employee requests across IT, HR, finance, facilities and other internal teams. An employee asks in plain language, usually in Slack or Microsoft Teams, and the agent verifies who they are, decides what needs to happen, and carries it out in the systems involved: it resets the MFA factor, grants the application access after the manager approves, files the time-off request, or starts the twelve onboarding tasks a new hire needs. People step in for approvals, exceptions and the requests the agent cannot safely finish. Classic ESM applies IT service management practices to other departments; agentic ESM keeps that shared service model and changes who performs the work.

ESM, AI-assisted ESM and agentic ESM

Enterprise service management is the practice of applying IT service management principles, a catalog, a front door, workflows, SLAs and a system of record, to departments beyond IT such as HR, facilities, legal and finance. It standardised how requests are handled. It did not reduce how many of them a person had to handle.

The first wave of AI in service management added assistants on top: a chatbot that answers from the knowledge base, a copilot that drafts a reply or summarises a ticket for the agent working it. Those tools are reactive. IBM's distinction is a useful test: AI assistants perform tasks at your request, while AI agents are proactive and autonomously plan and take actions to achieve a defined goal. Gartner draws the same line, describing AI assistants as dependent on human input rather than operating independently.

What makes a service desk agentic

The word is used loosely. Gartner coined the term "agent washing" for rebranding existing assistants, RPA and chatbots as agents, and estimated that only about 130 of the thousands of vendors claiming agentic AI are real. Five capabilities separate an agent from a chatbot with a new label.

1. It acts in your systems. The agent has scoped write access to the tools where requests are fulfilled: the identity provider (Okta, Entra ID, Google Workspace), the HRIS, MDM, EDR and business applications. An agent that can only answer and open tickets is an assistant.
2. It knows who is asking. Every action starts from a verified identity mapped to the employee record, so the agent knows the person's role, manager, devices, applications and location, and applies the right policy.
3. It plans multi-step work. Onboarding one person is a sequence across HR, IT and facilities. An agent decomposes the request, runs the steps in order, waits on approvals and recovers when a step fails.
4. It is governed. Which actions exist, who can request them and which need approval are defined by the organisation and enforced in code, not left to the model's judgement. Every action is logged.
5. It works across departments. One front door handles IT, HR and the rest, routing to the right team's policies and people, because employees do not sort their questions by org chart.

How agentic ESM works, step by step

Take a request that touches three departments: "My new hire Dana starts Monday and needs a laptop, Salesforce and a desk in the London office."

1. The agent identifies the requester in Slack or Teams and confirms they are Dana's manager in the HRIS.
2. It checks policy: the role's standard application bundle, the laptop model for the department, and whether Salesforce needs a licence approval.
3. It acts: creates or confirms Dana's account in the identity provider, assigns groups, requests the Salesforce licence approval from the budget owner, opens the hardware order, and raises the desk booking with facilities.
4. It waits for the approvals it needs, in the same conversation, and continues when they arrive.
5. It reports back with what is done and what is pending, and hands anything it cannot complete to the right human queue with the full context attached.

In a classic ESM setup the same request becomes three tickets in three queues, each waiting for someone to pick it up.

Benefits and risks

Benefits. Most employee requests are repetitive and policy-driven: access, passwords and MFA, onboarding and offboarding tasks, policy and benefits questions. When an agent resolves those end to end, employees get answers in seconds instead of hours, service teams spend their time on the requests that need judgement, and one platform replaces a portal per department.

Risks. Gartner predicted in June 2025 that over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls. The security risk is specific: OWASP lists "excessive agency", an agent with more functionality, permissions or autonomy than it needs, among the top risks for LLM applications, and its Top 10 for Agentic Applications adds tool misuse and identity and privilege abuse. The mitigations OWASP recommends map directly onto how an agentic service desk should be built: least-privilege actions, human approval for high-impact steps, and actions executed under the requester's own identity and permissions. The NIST AI Risk Management Framework is a reasonable structure for governing it.

How to evaluate an agentic ESM platform

Ask every vendor for evidence, not a roadmap:

• Resolution rate, defined. What share of requests do reference customers resolve with no human touch, and how is "resolved" measured?
• The action list. Which write actions ship for your identity provider, HRIS, MDM and top applications, and how are new ones added?
• Guardrails. Show where approvals and permissions are enforced, what the audit log records, and what happens when the agent is asked to do something it is not allowed to do.
• Departments covered today. Which teams beyond IT run on it in production, with their own policies and owners?
• System of record. Does the platform hold tickets, assets, applications and employee context itself, or does it sit on top of another ITSM tool?
• Time to value. What is live after 30 days, measured in resolved requests?

Who offers agentic ESM in 2026

Most service management vendors now describe agentic capabilities. Their own descriptions, as of October 2026:

For a broader comparison, see the best ESM platforms in 2026 and AI-native vs. bolted-on ITSM.

How Harmony approaches agentic ESM

Harmony was built as an agentic service desk rather than a ticketing system with AI added. Employees ask in Slack or Microsoft Teams; the agent resolves IT and HR requests by acting through 100+ integrations across identity, HRIS, MDM and EDR, using what Harmony knows about each employee's role, devices and applications from its asset and application management modules. Approvals are routed to managers in chat, every action is logged, and requests the agent cannot finish go to a person with the context attached. Cross-department work such as onboarding and offboarding runs as one flow across HR and IT, and HR service delivery runs on the same agent. We wrote separately about what an AI agent in your IT stack should never be able to do.

This guide is written and maintained by Harmony, which is one of the vendors listed. Details about other products come from their public pages, listed under Sources, and were last checked in October 2026. If something is out of date, email us and we will fix it.