Skip to main content

Nerdio is Microsoft’s 2026 Americas ISV Partner of the Year Get the story

Blog

IT process automation: how to reduce repetitive operations work

Learn how IT process automation cuts repetitive work across Windows 365, Microsoft Intune, and Azure Virtual Desktop with governed, auditable workflows.

Monday, 8:15 AM. Two session hosts show "unavailable" in the Azure portal after a domain trust hiccup, the Patch Tuesday golden image still needs rebuilding, 40 contractors start Wednesday, and the queue already holds a dozen password resets. With IT process automation, a person reviews exceptions instead of executing those repeatable tasks step by step.

This guide is for enterprise IT operations leaders and desktop platform owners managing service desk overload, rapid provisioning, and operations across Windows Cloud (Microsoft's umbrella term for Windows 365 and Azure Virtual Desktop).

How IT process automation differs from RPA, orchestration, and AIOps

The four terms describe automation at different layers of the stack:

  • IT process automation (ITPA) runs rule-based workflows at the API and infrastructure layer.
  • Robotic process automation (RPA) drives an application's user interface the way a person would.
  • Service orchestration coordinates multiple systems and dependency chains end to end.
  • Artificial intelligence for IT operations (AIOps) applies machine learning to monitoring data and predicts or detects anomalies.

ITPA itself sits outside Gartner's and Forrester's standalone category lists, and Gartner's closest markets are RPA and service orchestration and automation platforms.

The table below places each term against the layer it operates on, its typical trigger, and a Windows Cloud example drawn from Windows 365 and Azure Virtual Desktop.

Term

Operates at

Typical trigger

Windows Cloud example

RPA

Application UI

Recorded or programmed

A bot clicking through a legacy HR app to read a new-hire record

ITPA

API and infrastructure layer

Schedule, event, API call

An Azure Automation runbook that deallocates idle session hosts at 7 PM

Orchestration

Multiple systems and dependency chains

Event and dependency chains

A ServiceNow request that provisions a Cloud PC and closes itself when the user signs in

AIOps

Monitoring data and machine learning

Anomaly, prediction

Microsoft Intune Advanced Analytics flagging a change in device behavior before anyone opens a ticket

 

The line that matters for your estate runs between the first two rows. RPA emulates a person at the keyboard. PowerShell and Graph API calls act on the infrastructure directly, and those changes land in logs you already keep. Azure Activity Log captures role assignment changes, and Intune audit logs capture Intune actions.

With the vocabulary in place, the next question is what the manual version of this work costs today.

What repetitive operations work costs before you automate it

Enterprise service desks handle an average of 10,675 tickets a month, and 34% of support teams report volumes climbing year over year in HDI's State of Tech Support in 2025. MetricNet's ROI of Service and Support benchmarks price that work at roughly $22 per Level 1 ticket and $69 per desktop support ticket, and a session host that someone drains, removes, and rebuilds by hand sits closer to the desktop-support tier.

Hiring cannot close the gap. IDC projects in a 2026 report that by 2027, some 40% of IT operations roles will not be fillable in the G20, so the next decision is which tasks to hand to a workflow first.

Where to automate first in a Windows Cloud estate

Rank candidates on four dimensions. Volume tells you how often the task runs. Standardization tells you whether the same steps resolve it every time. Measurability tells you whether you can prove the impact. Impact tells you what a delay costs the business. Session host scaling scores high on all four and goes first. A quarterly application certification review scores high on impact and low on everything else, so it waits.

The table below applies that ranking to the workloads in a Windows Cloud estate. Each row names the process, the platform it runs on, what native Microsoft tooling handles, and what still needs a person or a management layer to close.

Process

Platform

Native Microsoft method

What still needs a person or a management layer

Cloud PC provisioning

Windows 365

License plus group assignment on a provisioning policy starts provisioning automatically

Right-sizing after deployment; bulk deprovision is Public Preview

App and script delivery before first sign-in

Windows 365, Intune

Autopilot device preparation: up to 25 managed apps under app assignment guidance, 10 PowerShell scripts during OOBE; Windows 365 Flex (formerly Frontline) policy: 10 apps, 10 scripts

Pre-provisioning and self-deploying modes unsupported

Proactive endpoint fixes

Intune

Remediations: detection and remediation script pairs

Windows Enterprise E3/E5 or VDA licensing under Remediations licensing guidance

Session host scaling

Azure Virtual Desktop

Auto-scaling plans

No CPU or memory trigger; no tool sharing

Golden image update and rollout

Azure Virtual Desktop

Azure Image Builder; session host update replaces VMs in batches

Sysprep and drain coordination

FSLogix profile container management

Azure Virtual Desktop

FSLogix profile containers

Scripted quota, monitoring, cleanup

Password resets

Entra ID

Self-service password reset

User enrollment campaign

Ticket routing and request fulfillment

ITSM plus all of the above

ITSM tool integration, the leading self-service automation method at 46% in Stonebranch's 2026 report

Workflow design

 

Session host scaling shows why the fourth column is important. Native auto-scaling runs on schedules with named daily phases and has no metric-based trigger inside the feature. Dynamic auto-scaling, generally available since July 9, 2026, can create and delete hosts, but only for pooled multi-session host pools with session host configuration. That difference determines whether the workflow adjusts existing capacity or changes the size of the host pool itself.

The older Azure Automation scaling script only manages existing hosts, and its jobs fail after three hours. Microsoft also recommends against using Intune to manage non-persistent session hosts, because deleting VMs leaves orphaned device records behind.

IT teams manage Azure Virtual Desktop through the Azure Portal, PowerShell, and Microsoft Intune, with each surface supporting different management tasks. These boundaries explain where an Azure Virtual Desktop management layer must add triggers or lifecycle handling, sometimes both. Nerdio Manager for Enterprise adds CPU and memory triggers on top of Microsoft's schedule model.

How to build automation that survives its own success

A workflow that ships fast and breaks without people noticing, costs more than the manual version it replaced. Three practices keep programs alive past the first quarter. Standardize the process before automating it, scope each rollout narrowly, and plan for the operator who inherits the exception.

Standardize the process before you script it

Standardization is the step teams skip most often, and skipping it is a documented cause of stalled programs. Some practitioners phase the work as stabilization, then standardization, then optimization.

Their guidance on exceptions is blunt: "Do not automate exceptions before you understand the normal path. If the workflow is unclear, automation will encode confusion and make remediation harder later."

Deloitte's Tech Trends 2026 attributes Gartner's prediction that 40% of agentic projects will fail by 2027 to the same cause. Organizations automate broken processes instead of redesigning operations.

Ship one workflow at a time

The number of tasks to automate on day one is often less than the team wants to. Google's SRE Workbook advises starting with human-backed interfaces before full automation and keeping toil under 50% of engineering time.

In a Windows Cloud estate that means one scripted task first, then a scheduled chain that patches the template VM and reimages the host pool.

Underneath every stage sit scripts and runbooks against Microsoft APIs. Azure Automation runs PowerShell runbooks on a schedule, and the Windows 365 Graph API (v1.0) exposes provisioning and resizing operations, which the Power Automate Windows 365 connector wraps in flows.

Plan for the operator who inherits the exception

Every automated workflow eventually hits a case its author did not anticipate, and the person holding the pager needs to be the one most prepared to handle it.

Lisanne Bainbridge named the risk of leaving this to chance in "Ironies of Automation". Once automation takes over routine work, "operators will not practice skills as part of their ongoing work."

Name the on-call operator before the workflow ships, document the exception paths, and schedule regular hands-on time with the underlying tasks so the person answering the alert can act on it. That plan also answers the two questions every automated action raises later, which are who authorized it and who can prove it afterward.

Governance for actions nobody clicks

Automated actions raise the same governance questions as human ones. Who authorized the change, who can prove it, and who catches it if something drifts.

Three building blocks answer those questions inside a Windows Cloud estate: least-privilege identities and audit records, a change-approval model that fits automated work, and the Azure and Intune primitives that enforce both.

Least privilege and audit records for machine actions

Every automated action needs an identity with least privilege and an audit record it writes itself. NIST SP 800-53 Rev. 5 sets this out for machines as well as people. Control AC-6 applies least privilege to users and system processes, and control CM-3 calls for automated mechanisms that document proposed changes and block them until the designated approvals arrive.

Together, those controls make identity scope and approval state part of the workflow instead of an after-the-fact review.

Match the change type to the approval path

Identity and audit records answer who and what. Change enablement answers whether the workflow should have run at all. Standard changes, per ITSM.tools, "often require little or no additional approval and are frequently automated," while normal changes go to a change advisory board.

Your scheduled scale-down is a standard change. A first-ever golden image rollout to production is not, however good the script is. ISO/IEC guidance adds the interrupt problem: "When an automated process starts, it is hard to interrupt, unless controls are built in that serve as tollgates for further automation steps."

Classify each workflow as standard or normal before it ships, route it to the matching approval path, and build in pause or approval tollgates at any step the team would want to stop mid-run.

Enforce it with Azure and Intune primitives

The frameworks map cleanly to controls Microsoft already ships. You can assign built-in Azure Virtual Desktop roles to a managed identity, so your scaling automation can hold Desktop Virtualization Power On Off Contributor and nothing broader.

Azure Activity Log logs role assignment changes and custom role definition changes, viewable for the past 90 days, and Intune auditing "is enabled for all customers. It can't be disabled."

For drift, Azure Policy's machine configuration feature offers Audit, ApplyAndMonitor, and ApplyAndAutoCorrect assignment types, and only the last corrects an unauthorized change without a person triggering remediation.

The controls' role is to make automated administration provable in support of a compliance strategy, with certification outside their scope. Native tooling leaves the operational layer above the primitives open, and a management platform earns its place there.

Where Nerdio Manager fits as an automation layer over Windows Cloud

Nerdio Manager is the automation layer that sits over the native tools this guide has covered. It deploys from Microsoft Marketplace into your Azure tenant, runs its automations through a service principal you control, and calls Microsoft Graph and Azure APIs to reach Windows 365, Intune, and Azure Virtual Desktop from one console. The Computacenter customer story shows what that looks like when a team runs both cloud desktop paths together.

These capabilities close the gaps the earlier tables named:

  • Patented auto-scaling: Evaluates CPU utilization, memory load, and session demand on each session host, provisions new hosts when demand exceeds the pool, and moves the OS disks of stopped hosts to a cheaper storage tier. TechTarget's Enterprise Strategy Group found organizations using Nerdio Manager with Azure Virtual Desktop saw up to a 55% reduction in Azure compute in its economic validation.
  • Auto-Heal: Handles unavailable session hosts without a human in the loop. During the auto-scaling run, Nerdio Manager can automatically attempt to repair "broken" session hosts by restarting and deleting/recreating them.
  • Desktop orchestration: Chains the monthly patch run from the template VM through to the host pool refresh. In independent testing by Dr. Benny Tritsch, updating a custom image took 146 clicks and 5:09 with native Azure Virtual Desktop management against 13 clicks and 0:37 in Nerdio Manager. Fewer manual steps also reduce configuration errors.
  • Scripted Actions: A library of PowerShell scripts and Azure runbooks that attach to schedules or run as Auto-Heal recovery actions, so the same fix covers routine execution and host recovery.
  • Unified Application Management: Deploys applications from one catalog to Windows 365 Cloud PCs enrolled in Intune and to Azure Virtual Desktop hosts in about 30 seconds, compared with up to three hours for native Intune delivery.
  • Extended Intune management: Nerdio Manager's primary Windows 365 value. It improves policy, baseline, and reporting workflows and supports creating Intune policies and protecting them through backup and restore.
  • Nerdio Advisor: Provides cost modeling, license guidance, and right-sizing recommendations for Windows 365 Cloud PCs and Azure Virtual Desktop resources.

Nerdio Manager governs these automations with RBAC and granular module permissions, agentless monitoring, and audit logging that shows who changed what.

Kevin Thompson, Senior Cloud Operations Manager at Sage, reported the headcount effect leadership asks about: "Before Nerdio, we were adding one head for every 40 customers we added. Now, we have grown from 200 to 1,000 customers without adding any additional headcount, saving 20-plus heads."

How to measure the program and when to add AI

Measure the program in units your CIO already tracks, and baseline before the first automation ships. Gartner reported in a 2024 press release that fewer than 20% of organizations have mastered measurement of hyperautomation initiatives, which is why many programs cannot defend their budget a year in. Five metrics keep the review honest:

  • Administrative hours per month on named tasks (image updates, host repairs, provisioning)
  • Azure compute cost per active user per month
  • Tickets per user per month, split by tier so deflection can be valued against the relevant support benchmark
  • Provisioning time per desktop, measured from request to first sign-in
  • Toil as a share of engineering time, against Google SRE's 50% ceiling

Once those numbers are stable, AI belongs above the deterministic layer, and the adoption data argues for caution. Forrester's Predictions 2026 expects fewer than 15% of firms to turn on the agentic features in their intelligent automation suites, and a 2025 study in the European Journal of Computer Science and Information Technology covering 12,487 incidents across 41 organizations found that self-healing systems resolved 71.3% of infrastructure-related incidents without human intervention.

Microsoft ships the entry points to build on. Intune Advanced Analytics includes anomaly detection, Copilot in Intune summarizes policies, and Nerdio Manager's auto-scaling intelligent recommendations run on dedicated Azure OpenAI resources in your tenant. Each entry point relies on the documented workflows and audit trails the earlier sections named.

On Monday morning, the unavailable session hosts are an Auto-Heal recovery action. The golden image rebuild is a scheduled workflow. The dozen password-reset tickets trigger an Entra self-service enrollment campaign, and the 40 contractors are a group assignment on a provisioning policy.

Get a demo to see how Nerdio Manager runs those workflows across your Windows 365 and Azure Virtual Desktop environment, or try it free in your Azure tenant.

Frequently asked questions about IT process automation

Ready to get started?