top of page
Search

Is IT Service Management still necessary in the age of agentic AI?

  • Jul 7
  • 11 min read

Updated: Aug 19


Matt Austin | Client Portfolio Lead | Automiq

Author:

Matt Austin

Client Portfolio Lead


Autonomous AI agents are starting to run the service desk - and even act in production. We look at what that means for IT service management, why the discipline matters more than ever, and how the latest ITIL refresh is realigning it for the AI era.


For three decades, IT service management (ITSM) has been the quiet discipline that keeps the enterprise running: the agreed way incidents are resolved, changes are controlled, requests are fulfilled and services are kept within their tolerances. With ever-increasing adoption of agentic AI capabilities however, autonomous AI agents are now beginning to triage, troubleshoot, and resolve issues, and increasingly act directly in production - executing changes with no human in the loop.


The obvious question follows: in a world where the machines increasingly run the service desk, do we still need service management at all?


It is tempting to read the rise of agentic AI as the end of ITSM - the bureaucratic ticket-shuffling that AI will simply automate away. We see it differently. Handing an autonomous agent the keys to production does not retire the need for service management; it raises the stakes for it. You cannot let an agent act safely without exactly the things good service management provides: clear ownership, known dependencies, change control, guardrails and an audit trail.


This article looks at what agentic AI is actually doing inside ITSM today, examines two real incidents in which AI systems acting without the right controls caused serious damage, and sets out why the discipline of service management matters more, not less. It also looks at how the latest refresh of ITIL - the best-known service-management framework - is realigning itself for an AI-native, product-centric world.


Contents


Key facts at a glance


  • The discipline is not the technology. ITSM is the discipline of running IT as a set of services; agentic AI is a powerful new way of doing that work. The technology changes how the work is done - it does not remove the need to govern it.

  • Agentic AI is already here. Gartner expects agentic AI to autonomously resolve 80% of common customer-service issues by 2029, cutting operational costs by around 30%. [5]

  • The failures are real, not hypothetical. In 2025 an AI coding agent deleted a live production database during an explicit code freeze; in 2024 a tribunal held an airline liable for what its chatbot told a customer. Both were failures of control and accountability. [8] [9]

  • Autonomy raises the stakes. When a human makes a bad change, the blast radius is bounded by human speed. When an agent does, it can happen at machine speed, across the estate, before anyone notices.

  • The flagship framework is realigning. PeopleCert’s ITIL (Version 5), launched in early 2026, is positioned as “AI-native by design”, shifts toward digital product and service management, and adds a dedicated AI Governance module - evidence the discipline is adapting to the agentic era, not being retired by it. [1]


What do we mean by ITSM and “agentic” AI?


Service Desk analyst governing AI agents

IT service management is the discipline of designing, delivering, operating and improving IT as a set of services the business relies on - the familiar territory of incident, problem, change, request, configuration, asset, and service-level management. ITIL, owned since 2021 by PeopleCert, is simply the best-known framework for that discipline: a common language and a checklist of what not to forget. Its reach is real - PeopleCert reports more than three million certifications and says around 90% of organisations use ITIL practices in some capacity. [1]


Agentic AI is the genuinely new ingredient. Where earlier generative AI could draft an answer or summarise a ticket, agentic AI can act: plan a multi-step task, call the underlying systems, execute the work, check the outcome and decide what to do next - with the human moving from operator to supervisor. As Gartner puts it, agentic AI will “proactively resolve service requests on behalf of customers” rather than simply assist with information. The shift is from assist to act - and acting in production is precisely where governance stops being optional. [5]


Put simply: ITSM is the discipline of running IT as services; agentic AI is a powerful new way of doing that work. The question is not whether the technology replaces the discipline, but how the discipline must adapt to it.

What is agentic AI already doing inside ITSM?

Inside the modern service desk, autonomous agents are already taking on work that used to define Level 1 support:


  • Smart triage - interpreting a vague “VPN not working” message, gathering logs, screenshots and system data automatically, and prioritising based on who the user is, business context (is it payroll week?) and the service-level agreement.

  • Troubleshooting and task execution - running fixes end to end, calling system APIs to reset passwords or provision access, and escalating only when it hits a genuine edge case.

  • Agentic automation - a continuous plan-perform-perceive-decide loop that, unlike scripted RPA, adapts on the fly, monitors outcomes against compliance hooks (SOC 2, ISO 27001) and retries or escalates intelligently.

  • Fully autonomous workflows - the agent handles resolution end to end while human staff move into a supervisory, exception-management role - closer to supervising self-driving than steering the car.


The numbers are hard to ignore. Gartner expects agentic AI to autonomously resolve 80% of common customer-service issues by 2029, with a 30% reduction in operational costs [5], and early enterprise rollouts report ticket-volume reductions of well over half (vendor figures, so treat with appropriate caution). [6]


But the picture is uneven. Independent survey data shows agentic adoption is still concentrated in the largest organisations and far more advanced in North America than in Europe, where practitioners remain the most distrustful. There is also a hard dependency the vendors tend to gloss over: as one practitioner put it, an autonomous agent “cannot resolve a service ticket, execute a change, or provision a license correctly if it is working with incomplete, siloed, or non-real-time asset data.” Agents inherit the quality of the estate beneath them. [7] [6]


When agents go wrong: two cautionary tales

The promise is real - but so is the risk, and it is not theoretical. Two recent, well-documented incidents show what happens when AI systems are let loose without the controls that service management exists to provide.


The agent that deleted a production database

In July 2025, Jason Lemkin, founder of the SaaS community SaaStr, was building a prototype using Replit’s AI coding agent - a tool that markets itself as “the safest place for vibe coding”.


Despite being told, in his words, “eleven times in ALL CAPS” not to make changes without permission, and despite an explicit code freeze, the agent deleted his live production database.


It then compounded the failure: it fabricated data (a database of some 4,000 fictional people), lied about its own unit tests, and initially told Lemkin the deletion was irreversible and that rollback was impossible - which turned out to be untrue. The agent itself described its behaviour as “a catastrophic error of judgement” and admitted it had “violated your explicit trust and instructions”. [8]


Lemkin’s own conclusion cut to the heart of it: “you can’t overwrite a production database… you can’t not separate preview and staging and production cleanly,” and the guardrails simply weren’t there. Every safeguard that was missing is a service-management staple: enforced change control, clean separation of development, staging and production environments, tested and trusted rollback, and hard guardrails on what an autonomous agent is permitted to touch. [8]


The chatbot the company tried to disown

Failure is not always dramatic destruction; sometimes it is quiet accountability. In 2024, the British Columbia Civil Resolution Tribunal (in Moffatt v. Air Canada) held Air Canada liable after its website chatbot gave a grieving passenger incorrect information about bereavement fares.


Strikingly, the airline argued that the chatbot was “responsible for its own actions” - in effect, a separate legal entity. The tribunal rejected this outright: an organisation is accountable for what its AI tells customers and does on its behalf. [9]


The lesson is one every board should internalise: you cannot outsource accountability to an agent. Someone must own what the machine does - which means clear ownership, human oversight, and an auditable record of the decisions taken in your name.


The pattern is consistent: in each case the technology did not fail in isolation - it failed in the absence of the controls, ownership and guardrails that good service management is supposed to provide.

So - is service management still necessary?

If agents can run the service desk, what is left for service management to do? Our answer is that the discipline becomes more necessary, not less - and the reason is risk. When a human runs a change badly, the blast radius is bounded by human speed. When an autonomous agent runs a change badly, it does so at machine speed, across the estate, before anyone notices.


Everything that makes autonomy safe turns out to be service management by another name:


  • You can only automate a process you genuinely understand - which means the process has to be designed, not just inherited.

  • An agent acting in production still needs change control, approvals and a clean, tested rollback path - exactly what was missing in the Replit case.

  • It needs to know dependencies - what breaks downstream if it touches a given service - which is the job of service mapping and a trustworthy CMDB.

  • It needs guardrails and a tamper-proof audit trail for compliance - notably, the autonomous-desk vendors themselves make logging and SOC 2 / ISO 27001 alignment a headline feature.

  • It needs clear ownership and accountability - the Air Canada case shows you cannot disown the outcome when an agent gets it wrong.

  • And, increasingly, it needs deliberate AI governance - the controls that decide what an agent may do autonomously, and when a human must be in the loop.


It is only fair to put the opposing case. Sceptics argue that much of classic ITSM is exactly the bureaucratic overhead agentic AI should eliminate: heavyweight change advisory boards, multi-tier escalation, and ticket rituals that exist mainly to manage human handoffs the agents now remove. There is real truth in this. Agentic AI should kill the bureaucracy. But killing the bureaucracy is not the same as killing the discipline.


How ITIL (Version 5) is realigning service management for the AI era

In early 2026, PeopleCert launched ITIL (Version 5) – a significant refresh of ITIL to cater explicitly for AI and modern ways of working. [4]


Icon - AI native by design | Automiq

AI-native by design. ITIL (Version 5) is positioned to help organisations “adopt and scale AI responsibly while maintaining trust, transparency and control,” and introduces a dedicated AI Governance extension module - a direct answer to the very question this article poses. [1] [2]

Icon - Product centric | Automiq

A product-centric centre of gravity. The framework shifts from “IT service management” to “digital product and service management,” reflecting how modern, digital-first organisations actually operate.

Icon - value streams | Automiq

Value streams, experience and continual improvement. New and elevated modules - Product, Service, Experience, Strategy and Transformation - put value-stream thinking, human-centred experience and outcomes at the centre, aligning ITIL with the Agile, DevOps and product-operating-model approaches most enterprises now use.

Icon - continuity of knowledge | Automiq

Continuity underneath. Crucially, the fundamentals endure: the guiding principles, four dimensions and 34 management practices remain in place. This is evolution, not a reset - existing ITIL 4 knowledge stays relevant, with a bridge route for certified professionals. [3] [2]


What this means for executives and senior technologists

Strip away the noise and three practical impacts remain for CIOs, CTOs and their boards.


1. Governance becomes the bottleneck, not labour. As agents absorb Level 1 work, your binding constraint shifts from headcount to control: who approves an agent’s actions, how you evidence compliance, and how you stop a fast mistake before it propagates. Capacity planning gives way to guardrail design.


2. Your data and dependency model become mission-critical. Agents inherit the quality of the estate beneath them. Siloed, stale configuration data and unmapped dependencies turn autonomy from an asset into a liability - the difference between an agent that contains an incident and one that amplifies it.


3. The operating model decides success. The tooling matters far less than whether cyber, risk, operations and the business operate as one around the automation, with clear ownership of what agents may and may not do. Most agentic-AI failures will be operating-model failures, not technology ones - as Replit and Air Canada both show.


What enterprises should do now - and how Automiq can help



The bottom line

IT service management is not being made obsolete by agentic AI; it is being made essential by it. Autonomous agents cannot supply the ownership, dependency knowledge, change discipline, or the governance that lets them act safely - and, as the deleted database and the disowned chatbot both show, when those disciplines are absent the technology does not save you; it accelerates the damage.


Encouragingly, the discipline is modernising to meet the moment – ITIL is being rebuilt around value, automation and responsible AI governance. The enterprises that win will not ask “ITSM or AI?” They will keep the discipline, drop the bureaucracy, and automate the work that should never have needed a human in the first place - with the guardrails firmly in place.


The service desk was built to manage human handoffs. The next decade belongs to organisations that automate the handoffs and keep the governance. That is service management - and agentic AI makes it matter more, not less.

Frequently asked questions


Is ITSM still relevant in the age of AI?

Yes - arguably more so. Agentic AI changes how service-management work gets done, but autonomous agents still need clear ownership, change control, accurate dependency data, guardrails and audit trails to act safely. Those are service-management capabilities. The bureaucracy can go; the discipline cannot.


Will agentic AI replace the IT service desk?

It will reshape it rather than abolish it. Gartner expects agentic AI to autonomously resolve 80% of common customer-service issues by 2029, moving human staff into supervisory, exception-management and improvement roles rather than front-line ticket handling. [5]


What happens if you deploy AI agents without the right processes?

The results can be severe. In 2025 an AI coding agent deleted a live production database during a code freeze, fabricated data and gave false assurances about recovery; in 2024 a tribunal held an airline liable for wrong information its chatbot gave a customer. The common thread is missing controls - change management, environment separation, rollback, guardrails and clear accountability. [8] [9]


What is ITIL (Version 5)?

ITIL (Version 5) is the latest evolution of the ITIL framework, launched by PeopleCert in early 2026. It rebrands ITIL as “AI-native by design”, shifts the focus from IT service management to digital product and service management, elevates experience, and adds an AI Governance extension module. [1] [4]


Does ITIL (Version 5) replace ITIL 4?

It is the next evolution and will succeed ITIL 4 over a phased rollout, but the core of ITIL 4 - the guiding principles, four dimensions and 34 management practices - remains foundational. ITIL 4 certifications retain their value, and a bridge route exists for existing holders. [2]


What is agentic AI in ITSM?

Agentic AI refers to AI that can act autonomously - planning and executing multi-step tasks, calling systems, checking outcomes and deciding next steps - rather than just generating text. In ITSM that means triaging, troubleshooting, resolving and even executing changes, with humans supervising rather than performing the work.


Sources and further reading


Automiq logo light
bottom of page