Beyond the hype surrounding LLMs, AI is already reshaping telco operations.
AI in Telecoms: A Strategic Imperative
In November 2022, OpenAI brought LLMs into the mainstream by releasing ChatGPT to the public, while Nvidia’s market cap surpassed $1T in 2023 and $3T in 2025. Artificial Intelligence (AI) is seen now as a no-brainer of every industry based on IT and even further, has become a geopolitical stake.
However, Artificial Intelligence in telecoms is not limited to LLM and is developing at own pace – some being industry specific, others more transversal. This paper will review the opportunities and challenges brought by AI for telecom operators.
To look beyond the hype, we will focus on the most used or popular forms of AI: machine learning, large language models and Agentic AI.
Operators invested a long time ago into data technologies. To give some examples, Orange identified data/AI as a strategic pillar in 2019 (Engage 2025), and Telefónica created in 2016 a specific data division, LUCA (now Telefónica Tech).
Key AI Technologies: Predictive, Generative, and Agentic
Predictive AI is the oldest and most mature family of AI in telecoms. It uses traditional statistical and machine-learning models to detect patterns and forecast events based on historical data. Its power resides in the fact that there is no imperative to discover or explicitly define a rule. The algorithm adapts to the ingested data. The prerequisite is to be able to gather, organize and label vast amounts of data..
It then acquires oracle-like capabilities in detecting weak signals such as emerging faults before outages or early fragile behaviors of future churning customers. It can also model complex technical systems such as radio-propagation or multi-faceted network planning combining technical and geo-marketing information.
Generative AI (GEN AI), embodied by large language models (LLMs), can create new content—text, code, or knowledge summaries based on a huge knowledge library (text, computer code). The prerequisite is to be able to create (through training) the model and to have the associated computing power (typical NVIDIA GPUs).
With GenAI, computers give a full illusion of speaking natural language. On the operational side, it can present technical situations in intelligible terms or interact with customers in a personal manner (as opposed to a rigid legacy chatbot), and produce personalized customer/email scripts. Not surprisingly, GSMA reports that 47% of AI implementations are developed in customer care.
GEN AI is also an outstanding way to summarize, produce or gather information which affects all company departments using a knowledge base (technical, marketing, legal, …), which can be in use in every part of the organization.
The emerging technology is Agentic AI, a system of autonomous agents that perceive, decide, and act toward defined goals. They aim to be an orchestration layer above processes. Agentic AI often uses LLMs as a reasoning engine and is enhanced with other modules (memory, action/retroaction loop …). This opens many opportunities in high-level automation, where for example agents will be able to coordinate across different domains (such as IT and customer care) and launch a consistent set of work orders for network maintenance.
Now let’s have a look on some concrete examples. According to NVIDia report state of AI in Telecommunications, the main use cases of AI are to be found in the context of Customer Experience, Network planning and Field Operations. Let’s take a few examples.
Increasing Customer Value
On the business side, Customer Value Management (CVM) is often considered as the primary use case for data AI. This comes from the fact that its business casing relies on a critical top line (usually very high margin) benefits: Customer Lifetime Value. CVM has evolved in sophistication, using more machine learning to determine the Next Best Actions or next based offer to the customer. Unlike static configuration, AI-based CVM can carry multi-objective optimization (revenue, churn, capacity, fairness, contact-fatigue). At Orange it counts for half of the AI use cases registered, also because it is the oldest.
Through the use of CVM, Operators have reported significant savings such as 2–5% ARPU uplift, up to 30–50% in campaign management costs.
LLM will come into the picture for customer interaction personalization.
AI-led Customer experience is then a combination of the right customer solicitation, with the optimized target value objective (churn or upsell), and the best interaction experience.
The next evolution lies in agentic CVM systems—autonomous marketing agents that monitor KPIs (NPS, ARPU, churn risk), learn from campaign outcomes, and autonomously trigger new actions.
For example, a customer-retention agent could, following an outage, and on the basis of customer scoring, coordinate with network-quality agents to deliver accurate customer communication as well as prioritized remediation, with an adapted loyalty offer—all without manual intervention.
Network Operations, Predictive Maintenance, Planning
On the technical side, networks generate billions of events daily. Predictive maintenance based on AI helps to predict probability of failures before they come into reality. This can generate significant savings in optimizing field operations.
All these operations help to improve KPIs such as the number of field interventions, mean time to repair, or Network Operating Center (NOC) agents’ productivity – which can be directly translated in financial benefits. AI Agents are already used in Orange NOC for mobile network monitoring. They aggregate network knowledge base, ticket information and technical procedures to support operations.
AI can be especially efficient in the combination of several constraints that can be difficult to model with determinist equations: for instance balancing coverage, throughput, and energy efficiency. For instance, AI is used in mobile Network planning prioritize deployment based on customer potential and not only on traffic. This use case known as “Smart Capex” is already deployed in several orange subsidiaries.
Beyond the AI hype and benefits several barriers need to be overcome to unleash the full AI potential.
AI is a transformative technology which must be deployed into a legacy organization. CVM projects for instance are primarily about Organizational change – which is broader than an IT project, requiring transversal approach and executive support. In the real life, the full potential may be as well constrained by the organization (siloed channels, outsourcing, …).
The fair measurement of AI benefits is also a challenge. Orange has developed its own methodology (Data Value Measurement) to compare the value of different projects. But in the end, there is a mixture of avoided loss (customer retention), avoided cost (operation optimization), and new revenues (upsell), which differ in budget nature – and in timeline.
The quality and labeling of data still need to be improved. Models trained on old network configurations may lose relevance—a phenomenon known as model drift. Hallucinations of LLMs are also a concern; Gen AI output can never be 100% trusted. These examples require regularly retraining models or manual checks, reducing automation benefits.
With neural networks and latent spaces of LLMs, AI systems have become black boxes. Their strength, which lies in the fact that no one must build complex laws to make them work, becomes a weakness when the time comes to explain an AI-led action. This can be a problem of control (do we let AI take decisions?), quality (was it the best decision?) and ethics (do we control Bias?).
World economic forum reports that up to 2/3rds of current working hours across functions (marketing, legal, HR, …) will be transformed by LLMs. For every company there is a challenge of LLM literacy in an ever-evolving technological context (for instance prompt engineering is not required for simple inference), where employees must adopt and assimilate LLMs on steady basis. There is also a risk that LLM generates their own working overhead, delivering far less optimization to these functions than advertised.
In conclusion, beyond the hype surrounding LLMs, AI is already reshaping telco operations and is now a foundational pillar for automated interaction and the emergence of autonomous networks.
