Enterprises are building “AI factories” that fuse private 5G, edge cloud, and modern data platforms to automate processes and unlock new business models. A recent session moderated by Abe Nejad of The Network Media Group (NMG) explored how outcome-driven architectures rooted in connectivity, compute, and governance can scale AI from pilots to production.
Speakers:
Watch the full interview.
The session framed an AI factory as a production system for intelligence and action. Its core layers include: data as raw material (quality, labeling, lineage), model/infrastructure (GPU platforms and MLOps), serving (getting outputs into workflows), and security & governance (identity, authorization, audit, and safety). Rather than tech-first deployments, the leaders underscored an outcome-back approach: define the customer or operational KPI, then engineer the data, connectivity, and application paths that deliver it.
Private 5G and edge cloud were positioned as the low-latency bridge between the physical world and AI decisions. Real-world examples, such as drone-based coastal surveillance streaming video to vision models in sovereign AI clouds, illustrate how cellular links and QoS assurances keep inferencing timely and mission-critical. Bringing compute closer to data (micro data centers at or near the enterprise edge) converts telemetry into real-time actions.
Challenges remain less about single tools and more about organization and scale. Many enterprises are still in discovery or PoC mode; moving to production requires a 3–5 year roadmap, integration with legacy systems, and clear guardrails. The conversation emphasized data security and trust (PII protection, authorized sources, in-transit safeguards), model quality (bias, hallucination mitigation), and standards/open APIs to avoid new silos.
The role of telcos extends beyond connectivity. The panel highlighted a path to co-develop AI factories with enterprises, reimagining networks as a distributed data-center fabric, pushing GPU capacity and storage to the edge, and exposing capabilities via open, developer-friendly platforms. Looking ahead, agentic AI was seen as a natural application layer on top of AI factories: multi-agent systems augmenting teams as “co-workers,” provided the underlying factory is scalable, governed, and performant.
“I personally believe all enterprises will become agentic enterprises over time, and most of the augmentation will be done by agents. That’s why I call it a journey from co-pilot to co-worker – and getting that right is super important.”
