Artificial intelligence is rapidly moving beyond cloud-based copilots and generic chatbots into the hands of frontline workers who keep global supply chains, retail operations, healthcare services and logistics networks running. Yet many organisations continue to face challenges around cloud dependency, unpredictable token-based costs, data privacy and unreliable connectivity, making it difficult to scale AI where it can deliver the greatest operational impact.
Zebra Technologies is taking a different approach by bringing AI directly to the edge through on-device intelligence, Small Language Models (SLMs) and tokenless AI designed specifically for frontline environments. Rather than replacing workers, the company’s AI strategy focuses on augmenting human decision-making with fast, secure and context-aware assistance that continues to operate even without an internet connection.
Leading this vision across Europe, the Middle East and Africa is Hozefa Saylawala, Senior Director, EMEA – Strategy, Products and Sales Engineering at Zebra Technologies. A technology industry veteran with more than two decades of experience, Hozefa is responsible for driving the regional strategy behind Zebra’s intelligent automation, connected frontline and asset visibility initiatives while overseeing product strategy, industry solutions, sales engineering and the company’s ISV ecosystem. Widely recognised for his expertise in the practical application of AI and machine vision to solve enterprise challenges, he regularly advises customers and partners on transforming frontline operations through intelligent technologies.
In this exclusive interview with CNME, Hozefa explains why the future of enterprise AI lies beyond the cloud, how on-device intelligence and tokenless AI can reduce costs while improving productivity, and why organisations should rethink the way they deploy AI across their frontline workforce.
What does Zebra mean by frontline-optimised AI, and how is it different from the broader enterprise AI narrative?
Before answering that, I’d like to share some exciting news. Zebra Technologies was recently recognised by The Wall Street Journal among the world’s top AI companies, alongside names such as NVIDIA, Intel, Alphabet and Microsoft. It is a proud moment for us and reflects the unique role Zebra plays in the AI ecosystem. To understand why, think of today’s AI industry as a modern-day gold rush with three key participants. The first are the tool makers that build the powerful chips powering AI, represented by companies such as NVIDIA and Intel. The second are the landowners, represented by cloud and AI platform providers such as Microsoft and Alphabet, which provide the infrastructure and intelligence. The third are the miners, and this is where Zebra fits in. We take AI into the field and put it to work in the physical world. While many technology companies focus on AI for digital environments such as office software, websites and cloud applications, the global economy runs on physical work, including warehouses, retail stores, transportation networks and hospitals.
Zebra enables AI where frontline work actually happens. Millions of frontline workers use Zebra handheld computers, scanners and sensors every day. By embedding AI directly into these devices, Zebra bridges the gap between digital intelligence and physical operations. A warehouse worker needs immediate answers such as where a shipment is located, how to process a return or what task to perform next. That is why Zebra develops highly specialised, task-specific AI. The AI is trained on industry-specific workflows, standard operating procedures (SOPs) and operational processes across retail, transportation, logistics and healthcare, making it practical, accurate and ready for deployment from day one.
Frontline-optimised AI is designed specifically for the fast-paced environment in which frontline workers operate. It delivers responses in milliseconds, works even without an internet connection, avoids recurring token-based usage fees and keeps sensitive enterprise data securely on the device rather than sending it to the cloud.
Why are large cloud-dependent AI models not always practical for frontline environments such as warehouses, retail floors and field operations?
There are three primary reasons, which include speed, connectivity, and cost. The first is speed, where the Cloud-based AI requires every request to travel to the cloud before a response is returned. Even a delay of a few seconds can affect productivity when a worker is assisting a customer, scanning inventory or fulfilling an order. The second is connectivity. Frontline workers often operate in warehouse basements, remote locations or areas with unreliable Wi-Fi or mobile coverage. Without connectivity, cloud AI becomes unavailable when it is needed most. The third is cost. Most cloud AI platforms charge according to token consumption. For organisations with thousands of frontline workers using AI throughout the day, these recurring costs can quickly become significant and difficult to predict. For these reasons, cloud-dependent AI is not always the most practical model for frontline operations.
How do Small Language Models (SLMs) bring intelligence directly onto handheld mobile devices used by frontline workers?
Most people are familiar with Large Language Models (LLMs), but Small Language Models (SLMs) are equally important for enterprise operations. An SLM is essentially a compact AI model that runs directly on a Zebra mobile computer instead of relying on a cloud data centre. This delivers two significant advantages.
First, it places intelligence directly into the worker’s hands. Instead of sending every request to the cloud, the AI processes information locally and delivers responses almost instantly.
Second, it is highly specialised. Rather than drawing information from the entire internet, the SLM is trained on an organisation’s own SOPs, manuals, product catalogues and operational procedures. As a result, it provides highly relevant and context-specific answers without being distracted by irrelevant information.
This combination of local processing and domain-specific knowledge makes SLMs especially valuable for frontline environments.
Can you explain Zebra’s Frontline AI portfolio, including AI Enablers, AI Blueprints and AI Companion, in simple terms?
We have intentionally designed our Frontline AI portfolio so organisations can adopt AI at different levels depending on where they are in their digital transformation journey. Think of it as three connected building blocks.
The first layer is AI Enablers. These are specialised AI capabilities that perform individual tasks such as reading multiple barcodes simultaneously, recognising products on shelves or identifying objects through computer vision. They serve as reusable AI components that developers can easily integrate into applications instead of building sophisticated AI models from scratch. They are essentially the building blocks of enterprise AI.
The second layer is AI Blueprints. These combine multiple AI Enablers, APIs and workflows into ready-made solutions for common frontline scenarios such as inventory management, warehouse operations and automated shelf auditing. Instead of starting with a blank page, organisations receive proven frameworks that can be customised for their own operations, significantly reducing implementation time while ensuring AI continues to improve through ongoing monitoring and optimisation.
The third layer is AI Companion, which is the capability frontline workers interact with directly. It is a conversational AI assistant that runs on Zebra mobile devices and allows employees to ask questions naturally, such as, “How do I process a return?” or “What is the procedure for this task?” AI Companion searches the organisation’s own manuals and operating procedures, provides immediate answers and can even launch the relevant application or workflow to guide the worker through the next step.
Together, AI Enablers, AI Blueprints and AI Companion create a complete AI ecosystem that supports developers, enterprise IT teams and frontline employees alike. This layered approach enables organisations to accelerate AI adoption while delivering practical, measurable business outcomes across frontline operations.
What is tokenless AI, and why should CIOs and IT leaders pay attention to it?
Tokenless AI simply means AI that does not charge organisations every time it is used. Most cloud-based AI platforms calculate costs based on tokens, with every prompt and response contributing to ongoing usage charges. For CIOs and IT leaders, this creates a significant challenge because AI costs become increasingly difficult to forecast as adoption expands across the organisation.
Tokenless AI addresses this by enabling AI to run locally on Zebra devices using Small Language Models.
This provides several important business benefits. It delivers predictable budgets because AI costs are known upfront instead of increasing with usage. It enables unlimited AI usage without concerns about token consumption. Most importantly, it improves return on investment by eliminating recurring cloud AI costs while still delivering powerful AI capabilities to frontline workers.
Although AI-enabled devices may require a higher upfront investment, organisations achieve stronger long-term value through lower operating costs, improved productivity and greater cost predictability.
How can tokenless and localised AI architectures help enterprises avoid unpredictable cloud and API usage costs?
By keeping AI tokenless and localised, enterprises can avoid the unpredictable costs associated with cloud processing and API usage.
The biggest advantage is that AI runs directly on the Zebra device instead of sending every request to a cloud server for processing. I often compare this to owning your own highway rather than paying a toll every time you drive. Once AI operates locally, there is effectively no toll booth.
The second advantage is a fixed-cost model. The primary investment is the AI-enabled hardware itself. Once an organisation deploys a Zebra AI-capable device, the on-device AI capabilities can be used without recurring token or API charges.
This fundamentally changes the economics of enterprise AI, shifting organisations from unpredictable operational expenditure (OpEx) to a more predictable capital investment (CapEx). It gives CIOs greater budget certainty while allowing employees to use AI as often as they need without worrying about escalating usage costs.
How does Zebra’s AI strategy help organisations reduce cloud costs and improve the total cost of ownership (TCO) of frontline mobile computing?
Procurement cost is often the first concern customers raise, particularly organisations that are highly price-sensitive. However, businesses should evaluate technology over its entire lifecycle rather than focusing solely on the initial purchase price.
A mobile device typically remains in service for three to five years. During that period, recurring cloud AI charges, software subscriptions and productivity losses can significantly exceed the original hardware investment.
Zebra’s AI strategy improves total cost of ownership in three important ways.
First, by running AI directly on the device, organisations significantly reduce or even eliminate cloud processing costs.
Second, they avoid recurring software and AI usage fees while enabling frontline workers to become more productive using the same device. This increases operational efficiency without introducing additional ongoing expenses.
Third, because AI continues to function offline, employees remain productive even during network outages or in locations with limited connectivity. Eliminating downtime translates directly into higher productivity and lower operating costs.
Taken together, these benefits deliver a stronger return on investment and a significantly lower total cost of ownership throughout the device’s lifecycle.
Why should organisations upgrade to AI-enabled hardware now instead of waiting for AI use cases to mature further?
Waiting comes with its own cost. Organisations that delay AI adoption risk losing productivity and operational efficiency every day, even though those costs are not always immediately visible on the balance sheet.
Today’s Zebra mobile computers are already equipped with dedicated AI processors designed to run on-device AI models efficiently. They provide the foundation for both current and future AI-driven workflows.
There are three compelling reasons to upgrade now.
First, organisations can immediately improve workforce productivity and operational accuracy using AI capabilities that are already available today.
Second, early adopters gain a competitive advantage by enabling frontline employees to make faster, more informed decisions while competitors continue to rely on traditional workflows.
Third, organisations future-proof their technology investments. AI-ready hardware is designed to support evolving AI capabilities without requiring frequent hardware refreshes, resulting in stronger ROI and a longer technology lifecycle.
Rather than waiting for future AI innovations, organisations can begin realising measurable business value today while ensuring they are prepared for the next generation of enterprise AI.
How do Zebra’s AI offerings empower frontline workers rather than replace them, and what impact could this have on enterprise mobility in the Middle East?
Zebra’s AI is designed to augment people, not replace them. It acts as a digital assistant that helps frontline workers perform their jobs more efficiently, accurately and confidently.
Instead of relying on guesswork, employees can ask questions in natural language and receive immediate, context-aware guidance based on company procedures, policies and workflows. This enables workers to make better decisions while reducing errors.
AI also accelerates routine tasks by proactively providing relevant information and recommending the next best action. Rather than spending valuable time searching for information, workers can focus on serving customers, managing inventory or completing operational tasks.
Another major advantage is breaking down language barriers. Many frontline workers operate in multilingual environments where standard operating procedures may be written in English even though employees speak different native languages. Zebra’s AI capabilities support real-time translation, allowing workers to ask questions and receive guidance in their preferred language while accessing the same enterprise knowledge base.
For organisations across the Middle East, this creates a more productive, better-informed and more inclusive frontline workforce. AI is not replacing employees; it is equipping them with intelligent tools that improve decision-making, enhance service quality and enable enterprise mobility at a much higher level.
Image Credit: Zebra Technologies
