If you are someone that doesn’t quite get the hype around AI-enabled smart glasses–like the Meta Ray-Ban Display glasses–one way to think about them is that they help solve what I call “AI’s Last Mile Challenge”.
What do I mean by that? Think back to when telecom companies laid down billions of dollars’ worth of fiber optic and copper cables to connect America. It was relatively easy to run those cables across miles of open land between cities. But the hard part was extending that infrastructure to each individual household. That bottleneck was known as the “last mile.”
AI has a similar problem. Training a model on oceans of general information is one thing. But delivering answers that are meaningful in the specific context of my life, my work, and my environmentis a whole other challenge.
Routine vs. Exception
Let’s take a sales example.
- Routine: A rep asks the AI, “What’s the price of our standard package?” The AI quickly pulls the list price and delivers it. Straightforward.
- Exception: Another rep, in the middle of a client call, asks, “This client has a unique configuration and a discount agreement from last year—what should I quote them?”
In the first case, AI can handle it easily. In the second, the answer depends on context buried in the CRM, past contracts, and company-specific rules. Without that “last mile” of information, the AI is stuck—or worse, it might give a wrong answer.
Helping My Neighbor
I also think about this issue every time I help my elderly neighbor with a tech issue. If she gets locked out of Facebook, what she wants to say is simple: “Why isn’t this working?”
But for AI to help, it needs to know:
- What device and operating system she’s on
- Whether her internet connection is active
- Her recent login attempts and error messages
- If Facebook flagged her account for suspicious activity
All of that information is essential—but invisible to her and way outside her ability to describe these answers to AI.
What’s in the Last Mile?
So what exactly is this “last mile” that AI struggles with? Part of the challenge is that it’s not one single thing. It’s a patchwork of context that lives in different places, often scattered and hard to reach.
Some of it lives in our own heads. We know what we mean, but we can’t always put it into words. Imagine trying to remember the name of a person you just met at a conference. You know their face, their job, and even the conversation you had, but the actual name alludes you. The knowledge is there, but you can’t articulate it.
Other pieces are in our digital footprint—emails, calendars, text threads, and all the productivity tools we live inside every day. That’s where the real-time commitments, updates, and reminders are, but most AI systems don’t automatically see that context.
Then there are the workplace systems: CRMs, Customer Support Systems, Knowledge Bases, Financial systems, etc. These are the places where critical information is stored—pricing details, contracts, compliance requirements—but they sit in silos that AI can’t easily tap into.
Context also shows up in live interactions—conversations happening in meetings, tasks being assigned in real time, schedules shifting on the fly. By the time you try to explain that to an AI, the moment has often already passed.
And finally, there’s our physical surroundings. The open document on your desk, the object you’re holding, the equipment you’re trying to operate. AI can’t see what’s right in front of you unless we find ways to bridge that gap.
Right now, the default solution is to push this burden back onto the user: “Please explain more. Please rewrite your prompt. Please give me more detail.” But that’s asking a lot. People don’t always know what’s relevant, or how to describe it. And when you’re stressed, you don’t have the patience to craft the perfect query—you just think, “Just fix it!”
How to Bridge the Gap
The real opportunity lies in automating how AI gathers that last mile of context. Smart glasses are one path forward: by capturing environmental and situational cues in real time, they can give AI a window into what’s happening right in front of us.
But glasses aren’t the only way. APIs and system integrations can quietly feed an AI the information locked inside calendars, CRMs, or productivity apps—data that would take a user far too long to describe.
Smarter user experiences can also play a role. Instead of forcing you to provide every detail upfront, an AI could recognize what it’s missing and ask clarifying questions, much like a colleague would. Sensors and wearables add another dimension, streaming real-world data that brings the AI closer to your lived experience.
And for people worried about privacy, secure data vaults may become the mechanism for selectively sharing just the right slice of personal context with the AI.
In short, bridging the gap doesn’t hinge on a single technology. It’s a combination of tools, integrations, and design choices that work together to give AI the information it needs without burdening the user.
Why It Matters at Work
This last mile issue is critically important for AI to thrive in the workplace. Imagine a factory floor where AI can monitor conditions in real time and flag potential hazards before they become accidents. Or picture an office where workers aren’t wasting time digging for documents because their AI coach already knows what task they’re on and serves up the next step.
For sales reps, bridging the last mile could mean walking into a client meeting armed not just with product knowledge, but with insights pulled from that client’s history and preferences. For customer service agents, it could mean responding to each situation with empathy, because the AI understands not only the policy but also the nuance of the customer’s circumstance.
When AI can close this gap, it stops being just a clever answering machine. It becomes a trusted partner—one that doesn’t just know a lot, but knows me: my role, my work, my situation. And that’s when AI truly starts to make work safer, smarter, and more human.
Looking Ahead
Bridging the last mile is what turns AI from impressive to genuinely useful. Training large models to answer general questions is one thing; enabling them to respond effectively in the context of real work and real life is another.
But unlike the telecom industry’s last mile, this one may not have a single, straightforward solution. System integrations are complex and expensive. Privacy concerns raise real questions about how much context people will—or should—be willing to share. And every workplace will have its own threshold for what “useful” really means.
That’s why the last mile of AI is less a finish line to be crossed and more a frontier to be explored. Some of the answers may come from technology like wearables or smart glasses. Others may come from better user experiences or from new approaches to data ownership and consent.
In the end, the last mile isn’t about making AI spectacular. It’s about making it practical. When we get there will depend not just on technical progress, but on the choices organizations and individuals make about how they want AI to fit into their work and their lives.
