# Was your dream to be a Test User?

Desktop changes still tested by hand? An agent runs the loop and sends back proof.

Last updated: 2026-09-03
Canonical: https://desk-ops.com/

Founder: 25 years • 100+ Projects • 20k users

Become an architect (again)

The boring desktop work runs itself. You get the architect work back.

Get a plan, an agent, and your first change is coached.

Yes. I'll join the waitlist

Azure VMware AWS OpenAI NetScaler

Skip Miller — Founder, Desk-Ops. View my background.

Desk Ops

August 2026

## Hi, I’m Skip Miller

25 years as a desktop architect. Now automating enterprise projects

To the architects still testing by hand: there’s a better way.

If you run network ops or virtual desktops, you know the loop: sign in, test, reset, repeat. A small change can take days of tedium.

So I built the Desk-Ops Agent, a secure desktop service that tests every change the way your users work — point, click, observe. Claude Code writes the “vibe-coded” planned backend or desktop change; Desk-Ops proves it on a real desktop. The deployment loop closes end to end automatically: you’re the architect again.

A desktop agent that tests Claude’s backend changes

It doesn't take long to get started but it could change your life.

— Skip Miller

Founder, Desk-Ops

“If your company staff are struggling with tedious user testing and mounting “technical debt” then Desk-Ops automation is the perfect choice.”

From prompt to proof

## How Desk-Ops works

Desktop architects should not have to spend days manually proving every change. You set the goal in Claude, research the current guidance, and approve the plan.

Desk-Ops then safely runs the work on a real test VM and repeats until the goal passes or a safety check stops the run.

- 01 Kick off in Claude. One prompt. Claude reads the plan, and nothing runs without your sign-off.
- 02 The agent runs it. On a real test VM: the 28-setting standard applied and read back, live.
- 03 Results in your portal. 28/28 verified posts to your dashboard the same day. Proof, not promises.
- 04 The receipt. A result report with the live capture, before-and-after stills, and pass/fail checks.

See our in-progress work on the [Web Portal](https://desk-ops.com/demo/), the [Desk-Ops Agent](https://desk-ops.com/docs/agent/user-guide/), and a real [Test Report](https://desk-ops.com/demo/reports/standard-enforcement-demo/).

Software and Professional Services

## Here is what you get with a Desk-Ops Pilot Project

We install a supervised agent on your test VM, run one real change from your backlog, and return the video, registry read-back, and change-approval report.

Here is the order: a written plan first, then the install, then your first real backlog item, then the report. The install runs over one outbound connection, no VPN, no inbound ports.

Here is an overview of a Manual vs Desk-Ops workflow:

Manual

Plan

weeks

Proof of concept

days

Manual item tests

endless

Deploy then triage

weeks

Desk-Ops

Plan

hours

Proof of concept

hours

Automated item tests

minutes

Deploy with proof

hours

Here are some of the benefits:

- Research - deeply - before it acts — across your documentation, knowledge base, and estate.
- Simplifies - Holds years of technical debt in context without losing the thread. Untangles complexity like only AI can.
- Plans, runs, tests, and refines - Make changes in one supervised loop.
- Treats change like code — written, approved, traceable, and proven.
- Reports - Returns technical evidence and an executive-ready report.

After the Pilot: You keep the agent

### When you put it to work, it's one flat monthly rate.

After the paid pilot, Desk-Ops stays installed on your test VM, ready for the next backlog item. You're on a flat monthly rate, month to month. Thirty days' notice either way, no auto-renew.

Every run goes on the invoice at zero, so testing never adds a line item. You sign off on a number, a date, and a scope. No current project, no charge. What comes back is planned, tested, and documented.

We think you’ll keep finding work for it and we intend to earn that on every run. Let's discuss a pilot project plan.

Get a plan, an agent, and your first change is coached.

Yes. I'll join the waitlist

install

### Done-with-you On Premise Install

Key Vault sandbox credential, agent installed on your test VM over the Azure Arc channel, restricted-port safe copy. You never hand over a production credential — the agent gets a sandboxed one, scoped to a copy of your environment.

plan

### A Written Plan for your first pilot

Before anything runs, you get a written PoC plan: goals stated as pass/fail tests, not adjectives. You read it, you approve it, then the agent starts.

control

### Claude Code Integration (yours)

You confirm the result from your own Claude Code seat, not our demo environment. If it passes, it passes on your machine, under your login, where you can check it yourself.

evidence

### Full Log and Video Test Results

We run at least one real item from your backlog and return the proof: video of the run, a registry read-back, and a change-approval report. A run that fails cannot be passed off as green: the harness re-checks the machine and does not take the agent's word for it. Not a canned scenario — your problem.

coaching (training and safety)

### We require sandboxing and safety

A named person at your company runs every change with our architect alongside during the coach period. They hold the keys and get shared office hours and ongoing coaching options.

human architect (optional)

### Planning, design, and execution

If you'd like enterprise-grade deployment combining the power of more than 25 years of experience, frontier AI models, and our desktop agent completing the "loop" for the best desktop and change automation money can buy today. This is what you'd request.

## Bring the change you're tired of hand-testing.

We'll write a concrete plan, deploy your agent, and prove your first change by day 14.

Get a plan, an agent, and your first change is coached.

Yes. I'll join the waitlist

Slots typically open within a few weeks.

Thanks!

## Skip Miller

Enterprise AI Architect

After 25 years and 100-plus production deployments, I care about one thing: changes that ship without breaking anyone’s morning.

For the last two years I’ve rebuilt that practice around a loop: Claude Code makes the change, a real desktop proves it, and the run repeats until it’s right. Desk-Ops is that loop pointed at the grind I know best: deploying, managing, and testing desktops, delivered like code.

### Enterprise background

- 25 years consulting
- 20,000+ virtual desktops
- 100+ production deployments
- Azure / Citrix DevOps expert
- Hybrid cloud architect

### Certifications

- Azure Administrator (AZ-104)
- Azure Virtual Desktop (AZ-140)
- Citrix CCA-V
- Previously: MCSE, Cisco CCNA

### Career highlights

Architected 100+ Citrix / AVD / NetScaler deployments: secure hybrid VDI for firms from 50 seats to 20,000.

Enabled 20,000+ users across healthcare, legal, manufacturing, and multi-tenant MSPs.

Tahoe Forest Hospital: Epic Citrix turnaround, medical VDI with App Layers and HA NetScalers.
