What an evaluation commits you to.

Forty-five to sixty minutes, and no obligation at the end of it.

Bring your path inventory, the failure moment you need to prevent, your traffic priorities and your platform constraints. Bring the operational owner, a platform or systems engineer, and a security advisor where one is relevant — the conversation is worth more with all three present than repeated three times.

You leave with a concise problem map, an honest fit assessment and a proposed demonstration or pilot boundary. If Atlas does not fit your problem, that is the useful outcome of the briefing and we will say so in it.

Controlled failure, against criteria you set beforehand.

We reproduce link loss and degradation against agreed success criteria — the dashboard, a link killed with video still running, the hosted console, the terrain simulator where it is relevant. The criteria are agreed before the demonstration, so the result is a measurement rather than a performance.

Remote or on site, on our equipment or in a representative setup of yours. What ends the step is a decision about whether the behaviour you saw addresses the failure moment you described.

Your compute, your links, your traffic — and no procurement to get there.

A software-only pilot runs on compatible compute and communication paths you already operate. No dedicated Atlas networking appliance is required to establish the product case, so a pilot does not have to wait behind a hardware purchase or a capital cycle.

Scope is agreed in writing before it starts: which nodes, which links, which traffic, how long, what counts as success and who measures it. That agreement is the point of the pilot — an unbounded trial produces opinions, not evidence.

Where the commercial relationship actually begins.

Interfaces, security boundaries, ownership, support and lifecycle responsibilities are defined here, alongside licensing. This is the first step with money in it, and the first with an executed agreement behind it.

Specifications, acceptance criteria, support and availability commitments exist only in that agreement — never in a slide, a web page or this one. The legal boundary, stated plainly

Nothing has to be dismantled to go into service.

The pilot runs the same software on the same compute as the deployment. What changes around it is the agreement and a signed licence file — not the binary, not the architecture and not the hardware you already bought.

You keep the measurements.

The path inventory, the failure analysis and the numbers from the pilot describe your network, not our product. They remain useful whatever you decide — including deciding against us. We would rather lose at the pilot than at your acceptance test.

Defined at integration, in the agreement.

Support, lifecycle responsibilities, ownership of interfaces and the security boundary are settled in the integration step and written into the executed agreement. Availability and acceptance commitments live there and nowhere else.

Widen the estate, not the contract every time.

Most programmes start with one bounded network — a vehicle type, a site, a fleet segment — and widen once the model is proven. The licence is sized to what you actually run, so growing the estate is a change of scale rather than a new evaluation.