As 5G networks mature and 6G appears on the horizon, mobile network operators (MNOs) are understandably eager to extract new value from their infrastructure investments. However, in the pursuit of monetization, too many are chasing the wrong kind of artificial intelligence (AI). Most of today’s high-profile telco AI initiatives focus on cost-cutting, replacing customer service representatives with chatbots powered by large language models (LLMs). While such tools are certainly flashy, their impact is primarily confined to the billing department, not the network.

At Digital Global Systems (DGS), we see a far more transformative opportunity — one that not only reduces operating expenses but also drives new service revenues. It is time to shift the conversation from human-to-machine interfaces to machine-to-machine intelligence, where AI enables next-generation applications through ultra-reliable low-latency communication (URLLC) and dynamic spectrum optimization. The future of AI in telecom is not in replacing call center agents — it is in re-architecting how the radio access network (RAN) adapts, performs and creates value at the physical layer.

REAL-TIME INTELLIGENCE FOR THE PHYSICAL LAYER

Modern wireless networks are increasingly software-defined and virtualized. This architectural flexibility opens the door to embedding AI directly into the signal processing chain, especially within the L1 (physical layer) and L2 layers of the RAN stack. These are not chatbot domains. They are mission-critical environments where microsecond decisions significantly impact the viability of applications such as autonomous vehicles, smart factories and augmented reality.

Most AI efforts in this space have relied on traditional channel modeling assumptions — using tapped delay line (TDL) or cluster delay line models based on the idea that the operator controls all RF signals in the environment. Under these models, embedded reference signals are utilized to train deep learning algorithms, thereby enhancing functions such as channel estimation, equalization and signal decoding.

However, in today’s congested, contested and increasingly shared spectrum environments, those assumptions break down. RF environments now include a diverse mix of unmanaged signals — from Wi-Fi and CBRS to military comms and private 5G — that introduce significant uncertainty and noise. Deep learning models trained in sanitized environments often struggle to generalize, resulting in degraded performance when it matters most.

THE FOUNDATION OF REVENUE-GENERATING AI

DGS’s patented RF awareness-based AI approach addresses this. Rather than assuming a clean spectrum, this technology actively senses, identifies and learns from the complete set of RF signals — both known and unknown — within a given environment. Using external or embedded sensors, the system extracts actionable intelligence from real-time spectrum conditions and feeds it into a hybrid machine learning framework that includes support vector machines, convolutional neural networks and domain-specific knowledge models.

The result is not just smarter RAN operations, it’s monetizable performance. DGS’s methodology enables MNOs to meet the stringent requirements of URLLC applications and dynamic spectrum sharing, unlocking new service opportunities across verticals. We are talking about deterministic latency in manufacturing automation, guaranteed throughput for drone traffic management and seamless handovers for immersive AR/VR experiences in stadiums and smart cities.

MONETIZATION THROUGH PERFORMANCE GUARANTEES

For example, LLMs might help a telco save a few dollars per customer on support calls. RF-aware machine-to-machineAI, by contrast, enables entirely new service classes that command premium pricing. Consider these examples:

• Smart factories require wireless networks that guarantee sub-10 ms latency with 99.999 percent reliability. RF-aware AI can continuously monitor and mitigate interference, ensuring that these SLAs are met, even in shared spectrum bands like the 3.5 GHz CBRS. That translates into revenue from industrial contracts, not just cost avoidance.
• Teleoperated robotics — from logistics to remote surgery — demands low jitter and high availability. Traditional AI models trained in lab environments cannot respond quickly enough to the unpredictable spectrum dynamics of real-world deployments. Machine-to-machineAI adapts in real time, optimizing transmission paths and prioritizing critical data flows.
• Public safety and defense communications benefit from DGS’s patented ability to detect and classify non-cooperative signals, including interference and potential threats, in near real time. That kind of capability is not just valuable, it is billable, especially in multi-agency or defense-contracted scenarios.

In these use cases, the value is not in reducing headcount but in creating performance guarantees that were previously impossible to meet. Those guarantees are the foundation of differentiated services and premium pricing.

FROM STATIC MODELS TO LIVE LEARNING MACHINES

Conventional channel modeling techniques fail in contested environments because they treat unknown signals as unstructured noise. DGS’s RF-aware system rejects this oversimplification. Instead, we utilize inferential signal processing and hybrid AI models to identify patterns in co-channel activity, infer potential interference sources and dynamically reallocate spectrum usage in real time.

This approach draws on a decade of patented innovation in:

• Signal detection and channelization by inference processing
• Drone signal recognition via pattern classification
• Blind signal classification using constant spectral component analysis
• Prediction of interference patterns to inform RAN configuration
• Dynamic optimization of RF environment sampling.

Critically, this intelligence can be embedded into the software functions of the radio unit, distributed unit (DU) or centralized unit (CU), or integrated into the RAN intelligent controller (RIC) or multi-access edge computing layers. The result is a self-optimizing network, not just in name, but in function. It is a network that does not rely on a central cloud to make decisions, but one that responds in the moment, at the edge, where performance makes or breaks the service.

REAL MACHINE LEARNING FOR REAL MACHINES

One of the most significant differences between our approach and the prevailing AI trends in telecom is this: We build systems for machines talking to machines, not machines impersonating humans.

LLMs are good at parsing language but lack the timing, determinism and explainability needed to operate in real-time, mission-critical environments. DGS’s RF-aware machine learning models are designed with physical layer constraints in mind. They prioritize computational efficiency, operate under low size, weight and power conditions and adapt to RF changes in milliseconds, not seconds.

Furthermore, our AI doesn’t just predict — it acts. By feeding learned environmental data back into the scheduling, resource allocation and modulation decisions of the RAN, we create a closed-loop optimization cycle. This is what true “machine intelligence” looks like: sensing, interpreting and executing within a dynamic physical system.

4G TO 6G: DEPLOYABLE NOW, SCALABLE TOMORROW

DGS’s RF awareness-based AI can be deployed today in existing 4G LTE and 5G O-RAN architectures using external sensors or software updates to existing RAN components. As network operators transition to vRAN and disaggregated architectures, our models become valuable, enabling performance enhancements across DU, CU and RIC functions without requiring new hardware.

At the same time, DGS’s work is laying the groundwork for 6G, where native support for shared spectrum, spectrum sensing and dynamic allocation will be critical. By developing and proving out these capabilities now, we ensure that MNOs can lead, not lag, when the 6G wave arrives.

CONCLUSION: THE REAL AI OPPORTUNITY IS IN THE AIR

The telecom industry is at an inflection point. It can continue to chase marginal gains from LLMs that reduce customer service costs, or it can embrace a more fundamental shift — one that enables new business models, new service guarantees and new revenue streams powered by M2M AI.

DGS RF awareness-based AI transforms the network from a static delivery mechanism into an intelligent, adaptive service platform. It is time to move beyond the hype and focus on where AI truly creates value: in the airwaves, optimizing the physical layer and enabling the next generation of reliable, low-latency wireless services.