TL;DR:

  • Smart glasses are pivoting from hardware products to AI subscription services — Meta, Snap and others are moving toward recurring revenue models where the AI assistant is the core product
  • That shift changes the enterprise risk profile: you’re not just deploying a device, you’re enrolling in a cloud AI service that processes what your employees see and hear
  • IT leaders should be asking about data processing locations, retention policies, third-party sharing, and what happens when the subscription lapses

For most of 2024 and 2025, smart glasses were a hardware story. Specs, battery life, weight, display quality. But something shifted this year. Meta has sold close to ten million pairs of its Ray-Ban smart glasses, and the conversation has moved on — from “will anyone wear these” to “how do we actually manage them in an enterprise environment.”

The trigger is subscription AI. Meta, Snap, and a growing list of others are building their long-term business model around AI features that require an ongoing cloud connection. Your glasses become a gateway to an AI assistant. The assistant sees what you see, hears what you hear, and responds. That’s the product. And that product has data implications that standard device procurement processes weren’t designed to evaluate.

What Changed and Why It Matters

Hardware has a manageable enterprise risk profile. You know where the device is, you can wipe it, you can restrict what apps it runs. AI subscription services are different. The value is created in the cloud, the data flows upstream to be processed, and the terms of that processing are governed by a subscription agreement that your legal team may or may not have reviewed.

Here’s the thing: most enterprise deployments of smart glasses today are in field service, logistics, and training — environments where employees are doing sensitive work. A technician repairing critical infrastructure, a nurse reviewing a patient handover, a warehouse worker scanning inventory. The AI assistant that “helps” these workers is also, necessarily, observing the operational environment they work in.

That’s not inherently a problem. But it does mean the question “where does that data go and who can see it” is no longer a theoretical concern — it’s a procurement question.

The Questions You Need to Ask Before You Buy

Where is audio and visual data processed? Some AI assistant features work entirely on-device. Most don’t. If audio queries are being transcribed in the cloud, you need to know which cloud region and whether that’s consistent with your data residency requirements. Snap’s enterprise Spectacles, for example, have different data handling than the consumer version — but you need to verify that in writing, not just take it as assumed.

What’s the default retention period for processed data? “We process your query and return a result” sounds ephemeral. In practice, many AI systems retain query logs for model improvement, quality assurance, and abuse detection. The enterprise question is whether your data — including queries made in operational settings — is used for training, and whether you can opt out.

What happens to the data if you cancel the subscription? This is easy to overlook. You buy 200 pairs of smart glasses for your field service team, deploy them, build workflows around the AI assistant, and then the subscription terms change or the vendor gets acquired. Are there deletion guarantees? Portability? A transition period? Hardware you own; subscriptions you rent, and rental terms change.

Who can the vendor share data with? AI services often have sub-processors — third-party services handling transcription, translation, or inference. Your data processing agreement needs to cover the full chain, not just the primary vendor. This is table stakes GDPR compliance for UK and EU organisations, but it’s often not at the front of people’s minds when they’re evaluating hardware.

The Snap Spectacles Example

Snap’s enterprise offering — currently around $2,195 per device with availability expanding to the UK this year — is an interesting case study. It’s positioned explicitly for enterprise use cases, with a clearer separation between enterprise data handling and consumer products than most smart glasses on the market. But even with enterprise-focused products, the specifics of AI data handling require review.

The Lens Studio developer ecosystem that underlies Snap’s AR platform is powerful, but it means third-party lenses — AR overlays created by developers outside Snap — can also be deployed on Spectacles. What access do those third-party lenses have to sensor data? What’s the review process? These are questions that matter in a corporate deployment in a way they don’t for a consumer product used for personal photography.

Practical Steps for Enterprise IT

Rather than avoiding smart glasses because of these concerns — the productivity and safety use cases are genuinely compelling — the answer is a structured evaluation process that treats AI-enabled devices as a special category.

Before any deployment:

Request a Data Processing Agreement specifically covering AI features. Don’t rely on the standard device DPA. The AI subscription component needs its own documentation covering what data is processed, where, for how long, and under what conditions it’s shared.

Test in a controlled environment first. Before deploying to production environments, deploy to a controlled test environment and observe what network traffic the devices generate. Your security team can do this with packet inspection. The results are often illuminating.

Define which use cases are in scope. Not every role that could benefit from smart glasses should get them immediately. Start with environments where the confidentiality stakes are lowest, build operational confidence, and expand from there. A warehouse picking operation has different data sensitivity than a legal due diligence team.

Clarify the MDM story. Most enterprise device management platforms are adding smart glasses support, but maturity varies. Android XR devices (running standard Android) integrate better with existing MDM infrastructure than proprietary platforms. If you need centralised policy enforcement — screen recording restrictions, app allowlists, remote wipe — verify that your current MDM can deliver this before you commit.

Fair Enough: The Convenience Case

To be honest, most employees who use smart glasses in workplace settings aren’t particularly worried about these questions. They want the tool to work, they appreciate the hands-free workflow, and “where is my voice query processed” is not what they’re thinking about while fixing a fault or navigating a warehouse.

That’s fine — it’s not their job to worry about it. It’s the job of IT and legal and procurement, and right now those teams are often playing catch-up because the devices landed in someone’s hands via an operational manager’s budget before IT policy had a chance to catch up.

Getting ahead of that means treating the next wave of smart glasses deployments — Android XR devices from Samsung, XREAL, and others arriving through 2026 — as an opportunity to set policy rather than retrofit it. The underlying technology is genuinely useful. The data governance around the AI subscription layer just needs the same rigour you’d apply to any cloud service your employees use. Because that’s exactly what it is.