Think about the nurse at a patient’s bedside, the driver who spends the whole day in the cab, or the associate answering questions on the retail floor. These are the people who stand to gain the most from AI, and they’re usually the last to get it.

That isn’t because frontline teams lack interest. Technology investment has always started with desk-based knowledge workers and worked its way out to the field later. With AI, there was also a practical barrier. Most AI tools needed a constant connection and a data center behind them, and that doesn’t match how frontline jobs work.

That barrier is starting to come down. At a recent Stratix customer workshop, Alex Kalish, Chief Strategy and Solutions Officer at Stratix, explained what has changed. He also laid out what has to happen before a frontline AI pilot can become something that still runs in your stores, hospitals, or trucks three years from now.

Why the Frontline AI Conversation is Different Now

Much of the technology behind AI isn’t new. Speech recognition, translation, and image processing have run on mobile devices for years. What changed in the last few years was generative AI, and those models were too large to run anywhere but the cloud. Every request had to go out to a server and come back.

Two developments are expanding what a mobile device can handle on its own.

• AI models are smaller and more focused
  Newer models are more efficient, and they’re very good at specific tasks. Frontline workers rarely need an AI that knows everything. They need one that handles the same repetitive tasks well, every shift.

• Devices are more capable
  More mobile devices now ship with dedicated AI processing, plus the memory and software support needed to run AI workloads locally. Running AI on the device is cheaper and faster, and it depends far less on connectivity. Not every feature will work offline. But a device can now process locally and then sync a record or update a system once the connection comes back. For a driver in a dead zone or a clinician in a patient’s home, that’s the difference between a tool they can use and one they can’t.

The opportunity is large because of how frontline work is structured. These jobs are repetitive and time-sensitive. They can’t be done remotely, and they often see high turnover. A small improvement to one task, repeated by thousands of people every day, adds up fast.

The Four Requirements for Frontline AI at Scale

Technology is a prerequisite, but it isn’t the whole story. Plenty of AI pilots are technically sound and still never reach production. In Alex’s experience, four things must line up.

1. Technology: The device you buy today must run the workflow of year three

If your organization plans a device refresh in the next 12 months, those devices will likely be in the field for three to five years. Whether your AI runs locally, in the cloud, or in a hybrid of the two, the hardware you choose now sets what you can deploy later. That makes device procurement a strategic decision, often with more people, different timelines, and new criteria involved.

A few factors to weigh:

• Lifespan
  The lowest-cost device isn’t a bad device. It just may not support the workflows you plan to run in year two or year three.

• OS and application versions
  Falling behind on versions is easy to overlook, and it can limit what you’re able to deploy in the field.

• Uptime
  You built a workflow your teams now rely on. A damaged device sitting in a drawer supports none of it.

• Consistency
  Not every device in the fleet has to be identical. Everyone doing the same job should be able to count on the same capabilities.

This isn’t an argument for buying the most expensive device on the market. It’s an argument for asking one more question during evaluation: does this device meet our requirements today, and will it still meet them three years from now? Not every role needs AI capability, so match the device to what each role will actually do.

2. Use case: Narrow and measurable beats big and ambitious

This is where a lot of AI budget gets wasted. The organizations getting the most value from AI aren’t always running the most ambitious programs. They’re running the most measurable ones.

Measurement doesn’t guarantee success. What it gives you is a way to spot the pilots that aren’t working, end them early, and put that effort into the ones that are. “Measurement allows you to kill the pilots that aren’t working fast enough, so you can go find the ones that actually are,” Alex explained.

A strong frontline use case is a narrow task with one owner and a clear metric. Broad, vague initiatives are hard to baseline, hard to assign, and hard to track. If you can’t say what a task costs you today in errors, minutes, or dollars, you won’t be able to tell whether AI is improving it.

Here’s what that looks like across just some of the industries Stratix supports:

• Retail and hospitality: A new associate gets a handheld with product lookup and procedure answers built in, so they can help customers sooner. The goal is shorter onboarding and faster productivity.
• Healthcare: A nurse speaks notes into a handheld, and AI turns them into a structured record. Clinical approval still applies, but there’s less back-and-forth, and the record can update once connectivity returns, even for clinicians working in patients’ homes.
• Transportation and logistics: A driver captures proof of delivery or damage in the cab or yard. AI can read the label, flag visible damage, and confirm the photo shows what it should, then sync when the device reconnects.
• Field service and utilities: A technician on site gets guided diagnostics and inspection support for the specific asset they’re working on.

Each example is aimed at a different outcome. Some drive productivity, others reduce risk. Know which one you’re measuring before you start.

3. Accountability: One owner for the result, one person behind every action

Accountability is the least exciting of the four requirements, which is why it’s so often overlooked. It works at two levels.

• Organizational
  One person needs to own the outcome by name. That doesn’t mean “IT and operations jointly,” and it doesn’t mean a committee. The program can be cross-functional, but one person has to answer for the result.

• Technical
  Frontline teams often work on shared devices, and that’s fine. The problem starts when everyone signs in with the same credentials, often from a sticker on the back of the charging cradle. AI raises the stakes. Once a tool can generate content or update a business record, you need to know who approved what. If every action shows up under one generic login, you can’t answer that question, and your auditors will eventually ask it.

4. Governance: Make the approved path the easier path

Device risk used to be fairly contained: a lost phone with email on it. Now frontline devices can carry AI tools your organization never approved. Many organizations worry about employees using unauthorized AI tools or unintentionally exposing sensitive business information through public AI platforms. Recent Stratix research found 63% of organizations have experienced a data compromise linked to so-called “shadow AI”.

Management tools that address this exist. The gap is usually that they aren’t configured correctly, or that people prefer an unapproved tool because nobody gave them a better option. When AI access goes through personal accounts on corporate devices, you lose visibility into what data is going where.

Alex frames this as a design problem as much as a discipline problem. “People are going to take the easier path,” he said. “We’ve got to make the right way to work actually the easier way to work.” Your teams won’t wait for policy to catch up, so the sanctioned workflow has to be at least as easy to use as the alternative.

What this looks like on Samsung Galaxy devices

The four requirements become more concrete when you apply them to the tools frontline teams already carry. Samsung’s Galaxy AI is a useful example, because it runs some features on the device and others in the cloud. That’s the same local-versus-cloud decision Alex described.

Rugged devices built for frontline work

Samsung’s Galaxy XCover7 Pro smartphone and Galaxy Tab Active5 Pro tablet are designed for high-intensity fieldwork in sectors like logistics, healthcare, and manufacturing, and they include AI features for everyday tasks. For a refresh decision, the question isn’t whether the device has AI. It’s whether the specific features each role needs are supported on the model you choose, because Galaxy AI feature availability varies by device model.

Narrow use cases you can measure. Several Galaxy AI features line up with the kinds of focused, repetitive tasks that make strong pilots:

• Talking with customers and patients across languages
  Call Assist and Interpreter support real-time translation for calls and face-to-face conversations in 24 supported languages. A retail associate or home health clinician could use this in place of waiting for a bilingual colleague. Measure it by time to resolve a request.

• Turning spoken notes into documentation
  Field teams can use voice recordings with Transcript Assist to build preliminary report drafts in set templates covering the type, time, place, and description of an incident. That maps directly to the field service and healthcare examples above. Measure it by documentation time per job.

• Finding answers without leaving the task
  Browsing Assist and Circle to Search help with fact-checking, summarization, and translation without breaking workflows, which fits the new-associate onboarding scenario.

Know what needs a connection

Not every feature runs locally. Transcript Assist requires a network connection and a Samsung account login, while Interpreter can run on-device with downloaded language packs. Map this out by role and site before deployment, especially for teams that work in dead zones.

Governance tools that keep the approved path in place

Samsung Knox gives IT teams control over AI settings, including the ability to disable cloud processing where needed, and it works with major UEM and EMM platforms, including Microsoft Intune, VMware, and Knox Suite. For cloud features, data isn’t retained by Samsung or Google or used for model training. Tools like these only help when they’re configured correctly. That’s where many organizations need support, and it’s the gap Alex pointed to.

What To Do Next: Focus on Sequence, Not Timeline

The order of these steps matters more than the exact dates.

• Start now
  Find out what the devices your teams use today can actually do, by role and by site, not just what the spec sheet says. At the same time, identify the workflows you want to improve and decide which parts need to run on the device, in the cloud, or in a hybrid of both. Get this done before your next refresh. Once hardware is deployed, it’s one of the hardest decisions to unwind.

• Then pilot
  Pick one use case and one metric, such as time per task or error rate. Establish a baseline and measure against it. In parallel, write an AI device policy that gives people a clear, approved path and accounts for every device type in your environment, including shared devices.

• Then scale
  Expand the workflows your pilots prove out, and plan your next refresh around what each role will actually need.

The goal is to avoid spending on a full refresh and only then learning that your use case doesn’t hold up. Pilot first, understand the results, and let them guide your technology decisions.

You Don’t Have to Predict the Future to Start

Almost every organization is being asked some version of “What’s our AI strategy?” There’s pressure to predict every capability on the horizon, and no one can do that. The good news is that you don’t need to. You can deploy a valuable frontline use case now with what’s available today.

Capable devices. A narrow use case. A single owner. An audit trail that shows who did what. An approved path that’s easier than the unapproved one. Get those in place, and you’ll be in a strong position, whatever comes next.

The technology creates the opportunity. Making sure frontline teams can rely on it every shift is the operational work that turns a pilot into an outcome.

Planning for AI on your frontline devices? Connect with a Stratix expert to talk through device readiness, use case selection, and governance for your environment.