Artificial intelligence is increasingly being introduced into healthcare workflows, but questions remain about where it can deliver practical benefits for frontline workers and where the technology is still being overhyped.
Pulse+IT spoke with Yogesh Kulkarni, VP, Product and Solution Strategy at US-based Zebra Technologies, which operates across Asia-Pacific, about the administrative and workflow burdens facing healthcare workers, the role of AI at the point of care and the safeguards health services should consider before deploying the technology.
1. There is enormous interest in AI in healthcare at the moment. Where is it genuinely useful for frontline healthcare workers today, rather than still being hype?
The most genuine value of AI today is found on the front line, where it can be applied to solve immediate, practical challenges rather than focusing on abstract future concepts. At Zebra, we view AI’s trajectory as a shift from reactive digital assistants that merely answer prompts to proactive agents that anticipate frontline needs and orchestrate workflows. Today’s tangible benefits lie in using AI to automate tasks and facilitate new, more efficient ways of working. This includes streamlining communication between systems and care teams, reducing time spent searching for supplies and critical medical assets, and automating documentation. By focusing AI on where the work is done, we translate insights into real-world action and transform how care is delivered at the bedside.
2. What are the biggest administrative or workflow burdens facing frontline healthcare workers that AI could realistically remove?
Frontline clinicians spend a disproportionate amount of time on administrative tasks that take them away from their most critical focus: the patient. The biggest burdens that AI can realistically remove are centered on manual documentation and fragmented communication. For example, using ambient listening and natural language processing, AI can capture clinician-dictated assessments and medication logging details by voice, eliminating the need to return to a nursing station to update records. Furthermore, by interconnecting workflows, AI can orchestrate complex cross-departmental workflows, such as the discharge sequence – coordinating transport, pharmacy handoffs, and documentation reconciliation the moment a clinician marks a patient ready for discharge. These are the critical operational junctures where intelligence can significantly boost safety and hospital throughput.
3. Can you give us a concrete example of a healthcare workflow where AI is already saving frontline staff time, and what evidence is there that it is making a measurable difference?
One example of a use case is leveraging ambient listening as an AI “superpower” for nurses. On the front line, a nurse can simply speak out loud their assessments, needs, or reminders, such as, “Remind me to turn the patient in 1402 in two hours.” The AI, running securely on their enterprise mobile device or a wearable, captures this, documents the task, and sets a reminder without the nurse ever having to break stride or touch a screen. This hands-free interaction not only reduces cognitive load but also improves adherence to care policies and frees the nurse to focus on the patient. The measurable difference is in reclaiming significant time needed for documentation per shift, significantly cutting down alert fatigue, and reducing missed care tasks to directly improve clinical continuity.
4. Healthcare organisations have spent years adding more applications, devices and alerts to clinical environments. Is there a risk AI simply becomes another layer of technology staff have to manage?
That is a critical concern, and it’s why our design philosophy is that frontline AI must be invisible. AI is only effective when it dissolves into the existing technology stack, rather than becoming another app to manage. Frontline workers don’t have time for another tool; they need their existing workflows to be smarter. We engineer AI to be intuitive and low-friction, holding ourselves to the standard that interfaces should be consumer-grade intuitive – clinicians shouldn’t require dedicated software training to benefit from AI assistance on the devices they already carry. By embedding AI into the Zebra devices clinicians already carry and making it context-aware, it augments their work seamlessly without adding another layer of complexity or screen fatigue.
5. What are the biggest barriers preventing hospitals from deploying frontline AI at scale – integration, legacy systems, data quality, cost or workforce acceptance?
While all those factors are relevant, the most significant barriers are ensuring seamless integration into existing clinical workflows and gaining workforce acceptance. If an AI tool disrupts workflow or requires extensive training, adoption will fail. Our approach is to involve clinical end-users in the design process to ensure the technology fits their daily reality. Another major barrier is the unpredictability of cost, bandwidth constraints and latency with cloud-only models. That’s why Zebra has adopted a hybrid AI strategy. We perform time-sensitive “execution AI” tasks on the device at the edge for low latency and privacy, while reserving cloud infrastructure for heavier “reasoning and knowledge AI”. This provides a cost-effective, scalable and responsive operational architecture that frontline workers can trust.
6. What safeguards should healthcare organisations have in place around privacy, cybersecurity, accuracy and accountability before putting AI into frontline workflows?
In healthcare, there’s no room for compromise when it comes to data integrity and patient safety. The first safeguard is ensuring rigorous data privacy and compliance with regional regulatory standards, from HIPAA to Australian Privacy Principles (APPs) and local health data governance frameworks. Our approach is to use on-device AI to process and de-identify all Personally Identifiable Information (PII) from captured data – like faces in an image or specific voices in an audio stream – before that information is ever passed to downstream workflows or the cloud. This ensures privacy is protected by design. Secondly, information must be deterministic, accurate and factual, as hallucinations are disastrous at the point of care. Finally, accountability means keeping the human in the loop; AI should augment and provide guidance, but the clinician remains the ultimate decision-maker.
7. With so many AI products and pilots emerging, what should health services be asking vendors before agreeing to another AI deployment?
Health services need to cut through the hype and ask questions that focus on frontline reality. First, “How does this solution integrate with our existing EHR and legacy systems, and does it require our staff to change their proven clinical workflow?” Second, “Is your AI designed to be invisible and intuitive, or will it require significant change management and training?” Third, “What is your architecture? Is it a hybrid edge-and-cloud model that ensures low latency and cost predictability, or are we locked into a variable cloud-only service?” Finally, and most importantly, “Can you demonstrate how your solution provides measurable operational value, not in a lab, but in the chaotic, real-world environment of frontline care?”
8. What healthcare AI use cases do you think are currently being overhyped or introduced before the technology is ready?
The most overhyped use cases are those that aim for fully autonomous clinical diagnosis and treatment without robust clinician oversight. While AI is a phenomenal tool for research and data analysis, expecting it to independently and safely diagnose patients at the bedside is premature and carries unacceptable risks. The real, immediate value lies in what we call “pragmatic augmentation” – taking the routine, mundane tasks that bog people down and giving them the tools to do their jobs better. The goal is to master operational workflows, reduce administrative burdens, and improve communication first. Building trust through reliable, everyday operational wins is how we create a lasting foundation for the future of AI in healthcare.
Add Comment