The agent can advise, but cannot complete the work.
Promising automation stays trapped in experiments and support chat.
ENTERPRISE AI · STRATEGIC BRIEF
An agent can complete a command and still leave a workstation silently broken. When software acts on a real PC, organizations need to constrain what it can touch, preserve working state, and verify the outcome.
Pinman gives Windows workstations a versioned memory of the configurations that matter—and gives people, automation, and AI agents a controlled, auditable, and reversible way to work with them.
FROM DEMONSTRATION TO DEPLOYMENT
An impressive agent demonstration is not the same thing as a production operating model. Before an agent works on a real Windows machine, an organization needs concrete answers.
These questions are larger than any one product. Pinman addresses a specific part of them: the state, change, recovery, and evidence at the Windows workstation where an agent’s action has consequences.
THE WORKSTATION BLIND SPOT
The configuration of a working PC is spread across application data, the Windows Registry, operating-system settings, and connected devices. A screenshot or successful command exit code does not show the complete operational state that matters.
Promising automation stays trapped in experiments and support chat.
Security and operations teams inherit risk they cannot easily observe or reverse.
Legible state · scoped authority · durable history · verification
WHAT PINMAN PROVIDES
Pinman models the selected configuration that makes a particular machine or workflow operate correctly. The same model is available through the browser UI, CLI, REST API, and MCP interface.
A Schema selects the files, Registry areas, Windows settings, and device settings that matter for a particular machine or workflow.
Captures, comparisons, and the Journal preserve what the selected configuration was, what changed, and the context around the work.
Named Pinman operations, Profiles, roles, and product permissions give automation a narrower surface than unrestricted access to the PC.
A workflow can preserve selected state before a change, apply an intended result, and compare afterward to see whether it actually worked.
THE MATERIAL IMPROVEMENT
The immediate value is not “more AI.” It is a better operational method for people and automation working on consequential Windows machines.
THE ASK
Parkerhill Technology is seeking organizations with a consequential Windows workstation problem and an interest in determining whether Pinman can create a measurable improvement.
The proposed next step is a paid, structured evaluation focused on one workstation, workflow, or machine class. Together, we define the baseline, model the relevant environment, test the Pinman-assisted approach, measure the results, and decide whether and how to deploy.
A SUCCESSFUL EVALUATION SHOULD PRODUCE
GO DEEPER
The PDF expands the argument, the operational comparison, and the proposed evaluation model in a portable format.