When to keep ConfigMgr task-sequence OSD next to Autopilot

When to Use This Guide

Use this guide when a team plans to retire Configuration Manager task-sequence OSD because co-management or Autopilot is now in use.

Typical mistakes:

• Treating operating-system provisioning as a co-management workload slider
• Replacing bare-metal recovery with an Autopilot profile that never sees the device
• Assuming Autopilot existing-devices can stand in for every wipe-and-reload path
• Costing the OSD role out before boot media, drivers, and BitLocker recovery have an owner

Provisioning is a separate architectural decision from the seven co-management workloads. Moving Windows Update or device configuration to Intune does not retire task sequences.

This article is documentation-reviewed. It does not include a tested hardware matrix or production cost figures.

Related reading: Autopilot v1 vs v2, Autopilot hardware-hash import failures, Intune vs ConfigMgr.

Scope and Assumptions

This guide compares four jobs that are often collapsed into one we-should-be-on-Autopilot sentence:

Primary sources:

• Introduction to operating system deployment in Configuration Manager
• Windows Autopilot for existing devices

What Autopilot does not replace by itself

Autopilot registers a device and applies an enrolment profile. It does not, on its own:

• Boot a machine that has no working OS and no OEM Autopilot registration
• Inject networking or storage drivers before Windows Setup
• Own BitLocker recovery keys from a previous ConfigMgr build
• Replace PXE, boot media, or offline driver libraries
• Move because a co-management workload slider changed

If the device must be rebuilt in a room with no OEM registration and no working OS, you still need an OSD path or an equivalent recovery image.

Step 1: Separate the four jobs

Write the requirement as four lines, not one project name.

1. Bare-metal. Unknown hardware, no hash in the tenant, or the disk is empty.
2. Existing-device refresh. Device is on the old OS and must be rebuilt in place.
3. New-device Autopilot. Hash is already registered and the ESP is the provisioning surface.
4. Recovery. Failed sequence, failed ESP, or BitLocker lockout.

Microsoft's Autopilot existing-devices workflow is an in-place conversion path. It is not the same as ConfigMgr bare-metal OSD, and it is not the same as a cloud-reset of an already enrolled device.

Step 2: Map the operational dependencies

For each job, name the owner of:

• Boot image or firmware boot method
• Networking and storage drivers
• Domain or Entra join
• Device registration / hardware hash
• Application and policy payload after the OS is up
• BitLocker key escrow
• Failed-build recovery

A ConfigMgr task sequence can carry drivers, applications, and BitLocker in one sequence. Autopilot splits those across registration, ESP, Win32 apps, and compliance. Both can work. They fail in different places.

If Autopilot hardware-hash import is already a live incident class in your estate, keep the OSD path until that import path is boring. See Autopilot hardware-hash import failures.

Step 3: Decide what to retain

Keep ConfigMgr OSD while any of these remain true:

• You still receive devices that are not Autopilot-registered by the OEM
• You need bare-metal recovery media that works without a cloud enrolment
• Driver or firmware steps still live only in a task sequence
• BitLocker recovery for the current fleet is tied to the ConfigMgr build process
• Existing-device Autopilot has not been proven on the hardware you refresh most often

You can still:

• Pilot Autopilot for new registered devices
• Move co-management workloads that are not provisioning
• Stop advertising the task sequence to collections that now use Autopilot only

Do not invent a saving figure for retiring the OSD role. If you need a cost comparison, label every assumption and keep it out of this page until those numbers are yours.

Step 4: Exit criteria for retiring an OSD path

Retire a specific task sequence only when all of these are documented:

• Every hardware model in that sequence has a tested Autopilot or OEM path
• Bare-metal recovery has an owner and a tested media or cloud-reset alternative
• Driver and firmware updates have an Intune or vendor path
• BitLocker escrow is verified in Entra or your chosen service, not only assumed
• Failed-provisioning recovery no longer depends on that sequence
• The collection advertisement is removed after the last pilot succeeds, not before

Leave a read-only copy of the sequence and boot media until the first recovery incident on the new path has been handled.

Step 5: Validation and rollback

Validation is per job, not per slogan:

• New registered devices complete Autopilot ESP on the target model
• Existing-device conversions follow Microsoft's published workflow, not an ad-hoc wipe
• Bare-metal recovery still boots on the models you have not moved
• A failed ESP does not strand the device without a rebuild path

Rollback:

• Re-advertise the last known-good task sequence to the recovery collection
• Do not flip co-management workloads as a substitute for restoring boot media
• Do not delete driver packages until the replacement path has survived one failed build

Sources

• Introduction to operating system deployment in Configuration Manager
• Windows Autopilot for existing devices
• Related AdminSignal guides: Autopilot v1 vs v2, Hardware-hash import, Configuration Manager hub