A workstation prepared by an administrator by hand takes half a day and is unique: a different set of updates, a different driver version, an application installed from a different source, a setting changed “because it was faster that way”. A year later a hundred workstations are a hundred configurations, and every ticket begins with the question of what is on this particular machine. Standardisation means the answer is always the same and comes from a definition, not from history.

The goal

Any workstation can be swapped for a new one, and after an hour the user carries on with the same applications, settings and data, with no administrator at the computer. This goal forces every decision below.

Four layers

The base image. A supported operating system version, drivers for the standard hardware models, the management tool’s agent, nothing else. The image is refreshed quarterly so that a year of update backlog is not installed at every deployment. The alternative to a custom image: the manufacturer’s image with automatic enrolment into management at first boot; less work, less control over what the manufacturer added.

Configuration. Everything that distinguishes a corporate workstation from a home one: domain join, the security baseline, network settings, certificates, browser and office suite configuration. Delivered through Group Policy or the management tool according to an agreed split of responsibility, never typed in by hand.

Applications. Packages in the management tool, assigned to roles, not to people: a core set for everyone, the accounting department’s set, the designers’ set. The role follows from group membership in the directory. A new employee in a department receives the department’s applications the moment they are added to the group. An application outside a package is an exception with an owner and a review date.

User data. Documents, desktop, application settings in the profile: redirected to a network location or synchronised with the cloud. Nothing lives on the workstation’s disk whose loss would hurt. This is the only layer that has to be restored when a computer is swapped, and it restores itself at logon.

Hardware

Two to three models on offer for users: basic, mobile, high-performance. Every new model enters the standard only after its drivers are added to the image and a test deployment succeeds. A purchase outside the standard is an exception justified in writing, because every additional model is a separate driver set to maintain and a separate way for something not to work.

The deployment flow

  1. Computer taken out of the box, connected to the network, switched on.
  2. Automatic enrolment into management based on the serial number registered at purchase.
  3. Image or manufacturer configuration, domain join, agent.
  4. Configuration and applications according to the role of the user who will be assigned.
  5. The user logs on, data synchronises, work begins.

The administrator does not sit at the computer. If they do, the process has a gap that needs closing, not working around.

Naming and inventory

The computer name follows a pattern: location, type, serial or sequential number. The inventory in the management tool is the single source of truth: model, user, role, deployment date, compliance state. A spreadsheet of computers is out of date the moment it is saved.

Measures

Time from unboxing to user readiness. Share of workstations compliant with the standard in the management tool. Number of workstations with applications outside a package. Number of tickets whose resolution was “reinstall”. The last number drops fastest, because a reinstall stops being half a day lost.

Exceptions

There will always be some: a station with software for a decade-old controller, a director’s laptop with an application nobody else uses. An exception has an owner, a justification, a review date, and is recorded in the inventory. An exception without a review date becomes a second standard.

Summary

Standardisation is not uniform hardware but a split of the workstation into layers, each with its own source: a base image, managed configuration, applications by role, data off the disk. Replacing a computer takes an hour because nothing on it is unique. A ticket starts with diagnosis, not with inventory, because the answer to “what is on this machine” is always the same.