Why AI will define the future of telecom

As the telecom industry transitions toward 6G, the scale and complexity of mobile networks are expected to grow exponentially. Traditional network management approaches will struggle to cope with these demands. The focus is therefore shifting toward intelligent, context-aware, adaptive, and self-optimising networks capable of proactively managing traffic, ensuring reliability, and responding instantly to rapidly changing environment.

In this evolving landscape, Artificial Intelligence (AI) and Machine Learning (ML) have emerged as transformative technologies. They enhance the reliability and efficiency of data transmission, enable intelligent routing, and support adaptive modulation techniques, allowing telecom networks to operate with unprecedented efficiency.

Causal AI for 5G

and 6G Networks

Causal AI represents an advanced machine-learning approach that goes beyond identifying simple correlations in data. Instead, it focuses on understanding cause-and-effect relationships, enabling networks to make more informed decisions.

In wireless communication, the first step is channel establishment, which depends on both the base station and the user equipment.

Because 6G networks will use higher frequency bands, signal path loss increases significantly, requiring the use of large antenna arrays. In traditional communication systems, a control signal (pilot signal) is transmitted first to estimate channel quality before actual data transmission begins.

However, with large antenna arrays and high mobility scenarios, pilot overhead becomes very high, leaving less bandwidth for actual data.

In 6G pilot overhead in communication is projected to be 80% whereas it is 20% up to 4G.

AI/ML can significantly reduce this overhead. By learning channel characteristics through data-driven models, AI can help establish the communication channel more efficiently and reduce the need for repeated pilot overhead.

On the network side, AI/ML models can perform Root Cause Analysis (RCA) to determine the exact reason behind performance issues.

For example, when interference increases, AI can distinguish whether it is caused by: Environmental degradation, or Passive Intermodulation (PIM) generated within network equipment. This capability enables faster troubleshooting and proactive network maintenance.

Improving wireless performance through intelligent signal understanding

The air interface in telecom networks is extremely dynamic. Causal AI can address several critical use cases:

1. Channel State Information (CSI) Prediction

Accurate CSI estimation is essential for efficient communication. During CSI compression, vendor-specific and telecom service provider (TSP) differences can create challenges. A causally aware auto-encoder that incorporates vendor and operator inputs can help model these differences more effectively.

2. Blockage Prediction

High-frequency signals used in 6G are easily blocked. AI models can predict potential blockages and adjust transmission parameters accordingly.

3. Channel Charting

Channel data can be used to infer user location without directly accessing location data. This allows networks to estimate how far a user is and how degraded the channel may be, enabling parameter adjustments for better transmission. This capability also supports the emerging concept of Integrated Sensing and Communication (ISAC) in 6G.

4. Intelligent Beamforming

Large antenna arrays used in 6G require searching across both horizontal and vertical beam directions, which increases complexity exponentially. AI enables smarter beamforming strategies, improving signal quality and reducing search complexity.

Computer Vision for Intelligent Transmission

Using visual intelligence to maintain stable high-frequency links

5G and 6G networks rely on millimetre-wave (mmWave) and terahertz (THz) bands to achieve extremely high data rates. However, these signals have low penetration capability and can easily be blocked by buildings, cars, rain and trees.

Even small movements of the user equipment can disrupt the signal path.

To address this challenge, vision sensors mounted on Small Base Stations (SBSs) can detect user positions within their field of view. Using computer vision (CV), the base station can track user movement and predict trajectory, allowing the system to dynamically adjust beam directions and maintain strong connections.

By extracting geometric information from images and combining it with network data, systems can also predict downlink rates and pre-emptively switch cell associations.

Computer Vision in Reconfigurable Intelligent Surfaces (RIS)

RISs are used to combat obstacles by reflecting, absorbing, diffracting, scattering electromagnetic waves in a controlled way to reach the receiver. They are software controlled. RIS is a two- dimensional engineered surface made up of many passive elements, each capable of altering the amplitude and/or phase of incoming signals. Since people and objects are always moving, static RIS settings are not enough. CV can detect obstacles, human locations or reflective surfaces in the environment. It can learn optimal RIS configuration using supervised or reinforcement learning. This information can help to dynamically reconfigure RIS to maximise channel quality.

CV in Semantic Communication

In Semantic Communication only the essential meaning of the object is transmitted. It saves bandwidth and time. Two distinct Compression Ratios (CRs) are used. High compression semantic encoding is used for static frames, as identified by CV sensing, and Low compression semantic encoding is used for the valuable scene frames. A Knowledge Distillation based approach is used that enables the two models with different CRs of semantics to learn from each other thereby enhancing the semantic quality of transmission.

Transforming Telecom Operations with Intelligence

Current Challenges in Telecom Management

From one point to the other point of the network, Multiple Technology equipment, Multiple Vendor equipment and Multiple Cloud scenarios (Edge, Regional, Central) may be existing. Real time performance demands will be there with respect to latency and bandwidth. Network data volumes will be huge especially in the case of video streaming or online gaming. Hence manual management of the network will be inefficient.

AI/ML in Management & Orchestration

AI/ML is key in managing cloudified telecom complexity. It enables automation & intelligent orchestration. It can do Anomaly detection & Fault Localisation, Failure prediction & Proactive maintenance, Traffic Prediction & Resource optimization and Intent based monitoring.

AI/ ML in customer domain

AI/ML can be used for customer churn prediction, customer segmentation and for offering Personalised Recommendations based on history of usage by the customers. It can be used for voice & chatbot automation and financial fraud detection as it happens.

Utilisation of AI in BSNL

BSNL is exploring the use of AI in the network management; customer satisfaction and experience; to analyse churn trends and reduce it. It is building SLM (Small Language Model), specifically for compliance, to see where the network is bad and attempt to remediate it. It has AI for automated call-centre services, deploying the chatbot ‘Vaani’ to address customer grievances. BSNL has launched ‘recharge expert service' on its website that can automatically predict and suggest tariff plans to its subscribers based on the analysis of their behaviour.

Wayforward

According to the Minister of Communications, the future of telecom will require AI-driven architectures that are agile, responsive, and continuously evolving. Such a transformation represents a shift from traditional telecom structures to faster, more flexible, and intelligent network systems, capable of meeting the demands of the digital era.

(The author is a former Advisor, Department of Telecommunications (DoT), Government of India)