Smart glasses are seeking a new route into everyday life. After years of consumer promises, Snap is presenting Specs as a professional tool for factories, retail and field service. Partnerships with Salesforce, Nvidia and Amazon Web Services turn the glasses into a gateway for visual recognition, AI agents and contextual instructions. The computer no longer remains on a desk; it overlays what the worker is looking at.
A technician can see a procedure beside the component being repaired, identify a part and ask for help without putting tools down. Retail staff can access availability and product information. A remote expert can follow a field intervention and annotate the shared space. These are situations in which augmented reality responds to a concrete limit: working while keeping hands and attention on the physical world.
AI makes the interface more powerful because it does not merely display prepared content. It can recognise objects, interpret a scene, listen to a question and propose the next step. Specs Intelligence promises continuity between perception and action. The device becomes a situated guide: it does not only know which manual to open, but tries to understand the problem in front of the user.
The training benefit is evident. A new employee can receive progressive instructions, compare a gesture with a standard and document an anomaly. Expert knowledge becomes available at the exact moment it is needed. Yet this also changes the value of skills: if every procedure becomes a visual prompt, who controls the system that defines the correct way to work?
The same device that assists can also measure. Time, errors, gaze direction, pauses and movements can become performance data. Surveillance may emerge without an explicit managerial decision, as a by-product of a platform designed for optimisation. Information needed to complete a task must therefore be separated from information retained to assess the person.
People in front of the glasses also need protection. Colleagues, customers and passers-by may not know whether a camera is recording, recognising or simply feeding an image to the system. A light indicator helps but cannot explain what data is analysed or retained. Workplaces need zones, functions and moments in which automated perception is clearly disabled.
The price of the new Specs—more than two thousand dollars—shows that Snap is not yet pursuing a universal object. It is targeting cases where saved time or fewer errors justify the cost. This enterprise phase may shape the rules later inherited by cheaper mass-market devices.
To be credible, glasses must also support accountability. They should show when a model is uncertain, reveal the source of an instruction and allow the worker to challenge a procedure. A visual suggestion placed over an object appears more authoritative than a sentence on a screen; it should therefore expose probability and limits rather than hide them.
AI glasses enter work promising to increase the ability to see. The real question is who sees through them: the technician receiving assistance, the company measuring the process or the platform collecting data about the world. If these perspectives are not separated and governed, assistance becomes wearable surveillance. If they are, augmented reality may finally find a purpose beyond the demo.