An AI agent that operates the software you already have, the way a person does. It reads the screen, clicks, types and checks the result. The connection and access are configured for the target machine. A person can take control, with action logging included in the workflow.
Your firm still gets its own written quote after the free review, and the final price is agreed in a signed agreement. These are starting points, so nothing in the quote is a surprise.
Paid discovery of one task
Credited in full against the build.
$1,500
One task in a web portal
Carrier quotes, county filings, supplier portals.
from $3,500
One task in a desktop program or remote desktop
Tested on your own data before it goes live.
from $7,500
Each further task in the same environment
$2,500 to $5,000
Care plan, optional
Somebody watching it, and fixes when your software changes.
from $400 a month
US dollars, October 2026. Monthly plans are month to month. Prices shown are starting points. Larger builds are quoted per project, and the final price and payment terms are set out in your signed agreement.
One person spends the whole day copying data from one system into another.
Our software is fifteen years old, it has no API, and the vendor is gone.
The RPA company quoted six figures and wanted six months.
We cannot install anything on the client's machine, it is their VDI.
Every month end somebody sits until midnight doing filings on a portal.
There is a job in most offices that nobody would describe as a job if you asked them to. Somebody opens two windows, reads a number out of one and types it into the other, four hundred times. They are careful, so it mostly goes right, and the mistakes surface a month later in a reconciliation.
AI staff take that job over. The agent looks at the screen the way a person does, finds the field, clicks it, types, reads back what happened, and moves on. Because it works through the screen, it can use your system whether or not it has an API, whether the vendor still exists, and whether it is a web portal, a Windows application from 2009, or a remote desktop on somebody else's network.
Two things make it safe to run on a real business's systems. Control is leased: if a person touches the mouse, the agent is locked out until control is handed back, so it can never fight a human for the keyboard. And every action is written down, so an entry from three weeks ago is a log line you can read.
The agent works between two windows. The lease is the block above it, and the log underneath is how a wrong entry three weeks ago becomes a line rather than an argument.
Drag the drawing sideways to see all of it.
A written quote before anything starts, and you watch it work on your own data before it goes live.
Show us the task once on a screen share. We time it, count the steps and the exceptions, and put a number on what it costs you a year. You keep the written page, and that number is usually the whole conversation.
After scope review, with an estimated price and schedule. If an API connection or a small change in process would serve you better, the proposal says so.
Your files, your portal, your exceptions, run beside the person who does the job now, before it goes live.
On a machine in your office, or on a spare PC you already have. The logs go where you can read them.
Most firms selling AI automation resell somebody else's tool. AI staff run on our own software, so the lease, the takeover and the action log are features of a product we build and maintain ourselves.
Remote desktop access and tools that let AI agents see and operate computers, with permission and human takeover.
What it proves: Computer control connected to an AI workflow.
Built on DeskVNC, our remote desktop client with an agent control plane and lease based human takeover, a Windows UI Automation server, screen reading with OCR plus mouse and keyboard control, and browser automation for web portals.
The details that shape the plan and the price, set out before you call so the quote holds no surprises.
The best candidates are repetitive tasks with clear rules: copying records between systems, portal filings, reconciliations. Where a task has a judgment call in the middle, the agent does the routine part and a person makes the call.
It works at the speed of your software, so we time the task on your own machines before we quote and the date in the proposal reflects it.
Anything that touches money or a filing gets a human check step, built in and priced in from the start.
When your software has a good API, we use it and the job costs less. Working through the screen is how we reach the systems that have nothing better.
Where we have a page written for your industry, the chip links to it.
Four ways we put AI to work in a firm, and the systems work behind them.
An AI phone line and a talking website guide that answer from your own documents, day or night, and pass the details to you.
Agents and integrations that take the retyping, the portal filings and the document intake off your staff.
The AI your staff want to use, on a machine in your office or in your own cloud account, with access by role.
A written plan for where AI pays off, a policy your staff can follow, and a senior lead to keep it on course.
Live streams built to stay on air when a camera dies or the internet drops.
Show our engineering team the workflow and the applications involved. We agree a scope, access requirements and a practical next step before implementation.