Renu Navale is vice president of edge AI at Intel.
A patient monitoring system, an ultrasound unit, an endoscopy device: each is a complex piece of technology, and each is also an edge device by definition, since it collects data at the point of care. That’s a good starting point for understanding why edge computing and AI are converging so quickly in healthcare, and why the implications reach from the bedside to the infrastructure of the cities we live in.
Keeping compute closer to the patient solves two problems at once: it keeps sensitive data from traveling to the cloud, and it cuts the latency between data collection and result delivery, giving the physician more time with the patient. The same logic is compressing timelines across billing, genomic analysis and the biopharma pipeline. The harder, less obvious work is what has to be true before AI can be trusted to act on that data at all.
The Shift From Insight To Action
What’s genuinely new in 2026 is a qualitative shift in what AI is trusted to do: from producing dashboards, alerts and summaries to coordinating multi-step clinical and operational workflows within defined guardrails, prioritizing what needs attention and escalating exceptions to the right person at the right time.
That shift is why governance can’t be an afterthought. When AI moves from advising to participating in a workflow, three things must change before deployment: accountability (who owns the workflow, where decisions remain human, how escalations work), auditability (can you reconstruct what the system saw, did and why) and controllability (can you pause, override or adjust it in real time). These require clinical governance decisions aligning the CIO, clinical leadership and the board. Handled well, governance lets organizations scale AI safely and repeatably.
Governing The Data, Not Just The Algorithm
There’s a second layer underneath all of this: governance of the data itself, meaning what systems can access it and how it moves across vendors while protecting patient privacy. Healthcare is further along on some layers than others. HL7 and FHIR have built a real foundation for EHR interoperability, and DICOM provides a mature model for imaging data exchange. Closer to the bedside, progress is more uneven: device data is often fragmented and hard to normalize across vendors at scale.
Defining clinical data standards belongs with providers, regulators and standards bodies, not any single vendor. But technology providers have real work to do enabling the edge platforms where this data is generated, making multi-vendor data easier to ingest, normalize, secure and expose in a governed way that health systems can integrate on their own terms.
Edge Versus Cloud Is The Wrong Question
A lot of the debate here still gets framed as edge versus cloud, but that misses the point. It’s a design question: where should each part of a workflow run? Privacy-sensitive and time-critical processing, or anything that must keep functioning when connectivity drops, stays close to where care happens. Model development, population analytics and interoperability reporting are better suited to the cloud. Organizations that scale AI successfully design seamless interoperability between the two and deliberately decide which workloads belong where.
Concrete Goals, Not Broad Ambitions
Organizations exploring these technologies tend to fall into one of two camps: skeptics who equate AI with consumer chatbots, and pragmatists who arrive with specific use cases and measurable KPIs already defined. The most productive path is the same for both: anchor early work in specific, measurable outcomes rather than broad technology goals, cutting latency from image capture to diagnosis, or modeling total cost of ownership for a device category. A well-scoped pilot that proves one metric builds the case for broader deployment far more effectively than a generalized demo.
For CIOs looking to move, the best place to start is a bounded workflow tied to a data source they already trust, often built around imaging, monitoring or diagnostic devices. The goal isn’t the biggest idea; it’s a use case where the data, ownership and escalation path are already clear. Before launching, define the governance model: who owns the workflow, what the system can do, how it’s monitored, and how a person steps in when needed.
Healthcare also carries compliance considerations unique to the industry, including FDA requirements for embedded medical device software. Organizations that lock into proprietary systems too early risk struggling to adopt better tools as the field advances; the goal is to prove value, not to run pilots for their own sake.
The opportunity for edge AI in healthcare is real. The organizations that capture it will start with a specific problem, build governance that lets AI operate safely within clinical workflows, and keep their systems open enough to grow with the technology.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
