Is Connectivity The Missing Layer In Your AI Strategy?

By Vera Miretsky, VP R&D, floLIVE

Recently, I had a fascinating conversation with Will Townsend, Founder of Lonestar Advisory and Research, who asked me why connectivity is so crucial to AIoT. You can watch the full discussion below, or continue reading for my perspective on what many enterprises are overlooking as they build AI-powered products

It’s true that when we talk about generative AI and agentic AI, the focus is almost always on compute. Organizations are investing heavily in AI infrastructure and products, and rightly so. But connectivity is often treated as an afterthought.

In reality, many companies are building impressive AI-enabled products while missing a foundational element. They focus on training models and deploying AI capabilities, but they forget that these systems ultimately depend on physical devices operating in the real world. Those devices are deployed in streets, vehicles, factories, and grids. They must connect securely, with low latency, and in full compliance with local regulations and data sovereignty requirements.

You cannot simply develop an intelligent product and then “add a SIM.” Connectivity is not a plug-in component. It is a core part of the architecture.

Considering Connectivity from the Outset

We are increasingly seeing enterprises come to us with well-developed AI strategies. They have designed their models, structured their data processing pipelines, and determined where inference will run. Only later do they begin to ask how the devices will connect, how traffic will be routed, and how data will remain compliant across borders.

By that stage, much of the architecture is already fixed.

Connectivity must be considered from the beginning. It influences where compute can sit, how data moves, how latency behaves, and how sovereignty requirements are enforced.

At floLIVE, we prepared for this shift years ago. We built a global infrastructure with local presence across multiple regions. That architecture allows us to create true local connectivity while maintaining global control.

In practical terms, this means we can deploy packet gateways and connectivity functions close to where the AI model operates. If a customer is running models in AWS, Oracle Cloud, or in a private data center, we can place connectivity components in the same environment. This alignment between compute and connectivity reduces latency, supports local IP addressing, and ensures compliance with in-country regulations.

The Challenges for Legacy Telecom Environments 

Many legacy telecom environments were not designed for this level of flexibility. Traditional architectures often centralize core functions or rely heavily on roaming models that introduce unpredictability in routing and performance. That approach may have worked for earlier generations of connectivity, but it creates constraints in modern AI deployments where data sovereignty, performance, and control are non-negotiable.

IoT connectivity itself presents unique challenges that are frequently underestimated. IoT traffic patterns differ from consumer mobile usage. Devices may generate continuous telemetry or burst traffic. They may operate in constrained environments. They may support mission-critical services such as emergency systems or grid management, where downtime is not acceptable.

To address this, we built a cloud-native, microservices-based core and connectivity management platform from the ground up. Because we own and develop all critical components, from the applet to the core network, connectivity management platform, charging, and billing, we have full control over how the system evolves. That ownership enables us to expand capacity, scale data processing, and introduce new capabilities without being constrained by third-party dependencies.

Providing Illumination

This architectural foundation becomes especially powerful when combined with real-time insights.

Every event across our network feeds into our systems. Over time, we have built structured data pipelines that collect and process large volumes of connectivity information. We train our own AI models on this data, enabling us to understand the behavioral patterns of individual devices.

Each device has a story. By analyzing historical behavior, we can detect deviations that may indicate a malfunction, performance degradation, or even a potential security issue. When behavior changes unexpectedly, we can generate alerts in real time and, importantly, take action. This is what we refer to as illumination.

Because we see the signaling, usage, and operational patterns across the network, we can provide customers with insights that go beyond simple connectivity status. We can help them operate fleets more effectively, predict issues before they escalate, and maintain availability in environments where reliability is critical. The results are truly amazing. 

Doing it All For Our Customers

Building such a system at scale is not trivial. We process millions of events per day, and delivering meaningful insight in real time requires constant investment in research and development. But for customers operating mission-critical services, the ability to monitor, respond, and control connectivity in real time directly impacts business outcomes.

AI is transforming industries. But intelligent models alone do not create intelligent systems. Those systems depend on devices, and devices depend on networks. If connectivity is not designed intentionally, performance, compliance, and scalability will eventually become constraints.

Enterprises that want to build the future with AI must ensure that connectivity is not an afterthought. It is part of the foundation.

Want to continue this conversation at MWC 2026? Let’s set a time to talk.