--- title: "Should you pilot XR glasses with an AI agent?" url: "https://vucale.com/blog/should-you-pilot-xr-glasses-with-an-ai-agent" description: "Display glasses with a built-in AI assistant are now priced and shipping. Before a pilot, check whether the glasses change anything in the process, and at what cost." --- # Should you pilot XR glasses with an AI agent? October 13, 2026·4 min read·Editorial team A young technician in a mint overall wears plain, unbranded glasses with a thin cable to a small unit in her pocket, checking a part by a shelf while a smiling colleague in a yellow sweater holds a clipboard A field-service manager watches a launch video. A technician looks at a machine, asks out loud what the error light means, and the answer appears in the corner of their glasses while both hands stay on the panel. The next morning there is a question on the leadership agenda: should we buy a few pairs and try this? The question is fair, because the products are now real. XREAL Aura, built with Google on Android XR, is the first priced display-glasses device on that platform: optical see-through glasses under 95 g, wired to a compute puck in the pocket, with Gemini as the built-in assistant and a camera with a privacy light for what the maker calls real-time AI perception. It starts at 1,279 USD, or 1,499 USD for the larger memory version, according to [XREAL’s product page](https://www.xreal.com/aura). Pre-orders opened on 7 October 2026 for people with an earlier reservation, and the first shipments go to the United States, Canada, Japan and South Korea. Other markets, Europe included, are promised for 2027. So the real question is not whether the device works. It is whether it changes anything in your process, and what you would have to prove before buying more than a handful. ## Why “let’s try it” goes wrong The glasses are an agent with a body. The assistant sees what the wearer sees, hears what they hear and can open apps on their behalf. That makes them more than a screen, and it also means that a pilot tests three things at once: the hardware, the assistant and the way people work. Most pilots only notice the first. **The pilot starts from the device, not the task.** Someone gets the glasses, looks for a use, and finds one that demos well. Months later nobody can say what improved, because nothing was measured before the glasses arrived. **Hands-free is assumed to be valuable.** It is, in some tasks. In many others the worker already has a free hand and a phone in their pocket, and the glasses add weight, a cable and a battery to manage, without saving a step. **The assistant is judged on a good day.** A demo shows the questions the assistant answers well. Work produces the ones it answers badly: the old machine without a manual online, the noisy hall, the part number in poor light. Those are the cases a pilot has to include. **Nobody asks where the data goes.** A camera that sees what the wearer sees also sees customers, colleagues, screens and documents. XREAL does not publicly set out which parts of the AI work run on the puck and which in the cloud. That is not an accusation, but it is a question you must have answered in writing before the glasses enter a customer’s site. **The availability is ignored.** For a European company the device cannot be bought through normal channels this year. A pilot built on imported units leaves you with no local support, no spare parts and a gap of months before you could scale, even if the pilot succeeds. ## How to diagnose it and design an honest pilot **Pick one task where eyes and hands are both busy.** Good candidates share three traits: the worker needs information during the task, not before it; stopping to look at a phone or a terminal costs time or adds risk; and the task repeats often enough to measure. Inspection with a checklist, guided assembly and picking in a warehouse fit. Writing a report afterwards does not. **Measure the baseline first.** Time per task, errors, how often someone has to call for help. Without it you will compare the glasses with a memory, and the memory always loses. We wrote about this in [measuring the baseline before an AI pilot](https://vucale.com/blog/measure-the-baseline-before-the-ai-pilot); it applies to hardware even more, because hardware is harder to give back. **Test the assistant and the glasses separately.** Run the same task with the AI assistant on a phone, and with the glasses. If the phone version captures most of the gain, the glasses are a comfort upgrade, not the source of value, and you can price them accordingly. **Write the privacy rules before the first day.** Where the camera may be on, who is told, what is recorded, how long anything is kept and who can see it. Tell the people who will be filmed, not only the wearer. Under EU data rules and the EU rules on artificial intelligence, a camera with an assistant behind it is a processing question, not a gadget question. **Count the whole cost.** The price of the glasses is the smallest part. Add device management, spares, charging, cleaning and sizing, the time of the pilot group and a fallback when the device fails mid-task. Ask the vendor what an enterprise fleet looks like; if the answer is vague, that is your answer for now. **Agree in advance what would make you stop.** A pilot with no stop condition only ever ends in expansion. Decide which result means “not yet” and write it down. The platform choice behind the glasses matters as much as the device itself. That is the subject of the companion piece on [Android XR and on-device agents](https://vucale.com/blog/android-xr-and-on-device-agents). As for us, we have no stake in XREAL, Google or any of their competitors. VUCALE, as the [homepage](https://vucale.com/) says, starts with diagnosis, and here the diagnosis can end with “do not buy yet”. ## A test you can run this week Pick one task you imagine the glasses helping with. Ask two people who do it to talk through the last time it went wrong: what they needed to know, when, and where their hands were at that moment. If the missing piece was information they could have read on a phone without stopping the work, start with the phone. If they needed it while both hands were busy and their eyes were on the work, you have a candidate for a glasses pilot — and a baseline story to measure it against. - XR glasses - AI agents - pilot - privacy ## Keep reading - [Agent factories will not stay in IT](https://vucale.com/blog/agent-factories-will-not-stay-in-it) - [The pilot worked. Scaling is a different project](https://vucale.com/blog/the-pilot-worked-scaling-is-a-different-project)