AI in Telecom: Current State and Trends to Watch in 2025

Gen AI entered 2024 with its fair share of skepticism. For example, 85% of executives were cautious to allocate significant budgets towards AI initiatives, with some experts calling it “a hammer looking for a problem”.

Despite skepticism, over the course of the last year AI in telecom moved from proof of concept into real deployments. Generative AI is rapidly transforming the telecommunication landscape in customer experience, network operations, and other niches.

Telcos are now leveraging AI-driven network planning, energy management, and edge computing to enhance customer engagement and bring data from the physical world closer to Gen AI to drive more automation. Besides, telco giants cooperate with Nvidia, today’s AI superpower, to research AI applications that will shape 6G—the future of wireless communications.

In this article, we’ll explore the latest AI-driven trends and innovations in the telecom industry, examining the real impact of AI and cutting through the hype to assess its current state.

Who’s at the Party? Key Players Driving AI in Telecom

Numerous established telecom giants and smaller forward-looking corporations are driving AI adoption in telecom.

Starting from the top players, let’s look at T-Mobile. In mid-September 2024, the company joined forces with OpenAI to develop IntentCX, an AI-powered platform that uses real-time data to understand and proactively meet customer needs. The system predicts customer intent, offering custom solutions, resolving issues, and even taking actions on behalf of customers. With access to billions of data points from customer interactions, the system is meant to maximize the success of every customer journey and provide instant, tailored support.

Comcast also announced AI-first initiatives meant to support network optimization and reliability. Janus, AI-enabled cloud-based network system, monitors network traffic patterns, predict and adapt to demand, adjust power use based on real-time network demand.

Ericsson, despite occupying a much smaller share of the telco market, is one of the leading innovators in the niche. Back in 2019 the company embedded AI throughout its 5G architecture, optimizing processes at every stage, from improving service quality to predicting hardware incidents. Fast forward to 2024, Ericsson launches Explainable AI, that identifies root causes of network issues, and suggests corrective actions. The system’s modular architecture allows for rapid deployment by other companies, accelerating AI adoption in telecom.

Zooming out, every large player in the global telecom market has some developments in AI that they’ve shared publicly:

AI-RAN Alliance: Bridging Business and Science for AI in Telecom

Besides developing their own solutions, telecom companies collaborate with universities and industry peers to accelerate 6G advancements and integrate AI to transform network capabilities.

Meet AI-RAN Alliance, a group of tech and telecom leaders that made it their mission to integrate AI into cellular technology and uncover the full potential of AI in Radio Access Networks (RAN).

The group has an impressive list of members, featuring Nvidia, Microsoft, T-Mobile, Samsung, Ericsson, and more. The alliance also hosts scientists and researchers from the leading universities around the world.

In a recent interview, Alex Jinsung Choi, Chair of the AI-RAN Alliance talked about three main working groups of the alliance and what they’re actually doing:

• AI for RAN: Focuses on using AI and ML to enhance RAN performance, including tasks like spectrum and traffic management and energy optimization.
• AI and RAN: Develops a unified computing platform that can handle both AI and RAN workloads, with an initial focus on GPU-based solutions for high computational needs.
• AI on RAN: Examines how to effectively run AI applications, such as Edge AI and real-time inferencing, directly on RAN infrastructure to support low-latency applications like autonomous driving and smart cities.

While work is still in progress, the alliance’s vector of activities gives us a hint where AI in telecom will be applied in the upcoming years.

Key Applications of AI in Telecommunications

Predictive Maintenance

One of the most crucial applications of AI in telecommunications is predictive maintenance. By bridging the gap between data captured by physical sensors and AI, companies can predict potential failures and adjust maintenance schedules accordingly.

Jeff Winter, industry 4.0 thought leader, and former Industry Executive at Microsoft believes that:

“Predictive maintenance can reduce maintenance costs by an average of 30% and reduce spare parts related costs by an average of 10%. It enables the optimization of maintenance schedules that leads to an average 40% reduction in unplanned downtime”.

With telecom experiencing the fastest-growing rate of outages among tech industries and downtime costing $33,333 per minute, companies are keen to use AI for maintenance solutions.

Leading telecom companies like Verizon and AT&T have embraced this technology to improve their infrastructure reliability and reduce repair costs. For instance, these companies have reduced outages by employing predictive models, significantly enhancing network availability.

Potential outcomes:

• AI in telecommunications can save up to 40% on maintenance and spare parts costs
• Predictive maintenance can reduce unexpected downtime by 40%, potentially saving telecom companies around $13,333 per minute in avoided downtime costs.

Network slicing and AI-based slice management

With the advance of 5G networks, telecom companies tapped into a $200B market of virtual networks. Network slicing is a feature of 5G networks that allows communication service providers (CSPs) create and customize multiple networks ****on a single physical network. Each network, or “slice”, can be customized to meet the specific requirements of consumers and businesses. For example:

• Low latency networks for mobile gaming (B2C) or autonomous driving (B2B)
• High data throughput networks for 4K and 8K streaming services
• High-performance networks for cloud vendors to deliver seamless experience to end consumers (B2B2C)

In the nearest future AI in telecom will play a key role in network slicing as it helps monitor and orchestrate virtual networks. Ericsson case study provides deeper explanation into why AI-enabled networks are superior to the conventional ones:

• SLA monitoring: AI algorithms monitor performance KPIs across slices to meet service level agreements. Machine learning models predict potential KPI degradations and suggest or automatically apply corrective actions, ensuring high service quality.
• Self-sustaining networks: AI predicts network issues and proactively adjusts resources, ensuring each slice maintains its required performance levels. This allows for real-time, self-adjusting networks that meet diverse service needs without manual intervention and human-in-the-loop.

Potential outcomes:

• AI in telecom allows to streamline virtual network management
• This helps CSPs expand their service offering and tap into $200B 5G slicing market

Network Optimization

A recent study found that 70% of telecom companies focus on improving their networks' reliability, with AI in telecom seen as an essential solution.

Currently, network planning and optimization (NPO) teams focus on improving radio coverage, capacity, and end-user perceived quality. While companies are only investigating practical applications of AI in telecommunication network optimization, there are several areas where GenAI could be beneficial.

• Prediction of traffic patterns: AI analyzes historical data and current network conditions to predict traffic spikes and improve quality of service during peak times.
• Automated interference detection: AI algorithms can distinguish normal traffic from interference signals to identify sources of interference faster. Here’s a case study by Vecta Labs about the first successful RF interference hunting.
• AI energy management: To improve energy efficiency and address climate concerns, telcos are using AI to monitor and optimize energy use across networks. For instance, Nokia’s AVA platform claims to deliver 2-5 times more power savings compared to traditional energy management systems.

However, GenAI also brings certain risks. Faris Mismar, Senior Principal Consultant of Wireless and AI at Nokia Bell Labs highlights the following concerns:

“There is no doubt that just as Generative AI comes with benefits, it also poses certain challenges to its adoptability in NPO such as current ways of working (including the measuring of resource utilization) and Generative AI challenges (such as hallucinations). “

Potential outcomes:

• AI-based RF interference detection can reduce manual analysis needs by up to 90%.
• Nokia reports its AI technology can save up to 30% in energy for telecom radio networks.
• Network optimization efforts also boost equipment availability and improve overall network efficiency.

Virtual Assistants and Chatbots

In telecom, every point in customer satisfaction (CSAT) is hard-earned, as even minor service issues are often met with strong negative reactions. Although telecom CSAT scores have generally improved globally, one area dropped below 70 points in 2024: call center satisfaction.

Today, customers hate waiting on hold to reach an available operator, especially during major outages when phone lines get busy. That’s why telecom companies develop AI assistants that replace first line support.

Chatbots come with multiple benefits, including:

• Instant support: Chatbots can resolve repetitive issues, help customers change their subscription plans, or provide updates about maintenance time instantly. This helps companies decrease Time To Resolution (TTR) metric, increasing overall customer satisfaction.
• Round-the-clock availability: Chatbots can answer questions any time, increasing customer satisfaction and reducing support expenses.
• Ultimate knowledge database: Unlike humans, AI-powered chatbots can memorize all information they’re trained on. This ensures that every question will be answered or directed to a support agent.

As for the real-world examples, Vodafone uses TOBi, an AI-powered virtual assistant that handles customer queries with machine-like efficiency, ensuring precise answers while escalating complex problems to human agents. This application of AI in telecommunication allows companies to automate customer service effectively without compromising the trust or preferences of their users.

Potential outcomes:

• Increased customer satisfaction, by lowering time to issue resolution
• Improved first call resolution rate
• Reduced support expenses
• Removing traditional phone support centers (especially relevant for Millennials and Gen Z clients)

Fraud Detection and Prevention

AI in telecommunications can help prevent spam calls, SMS phishing, SIM card clonning, and more.

In 2023, telecom companies lost around $39 billion dollars as a result of fraudulent activities, representing 2.5% of global telecom revenue. Robocall fraud alone costs roughly $2 billion dollars to telcos, not to mention reputation risks.

With the help of historical datasets and machine learning algorithms AI helps solve this multi-billion dollar problem. For example, Bharti Airtel claims their AI-powered detection system identified 154 million potential spam calls and 8 million spam SMS within just one month.

Some developments come even further and detect fraudulent activities based on call records and user behaviors. However, privacy issues arise because such monitoring often requires access to user communications.

Orange is another company that successfully implemented AI in telecom for fraud detection. The AI model analyzes 100-160 million call records daily to identify fraud patterns. Steve Jarrett, Chief AI Officer at Orange says: “We can identify fraud patterns very rapidly. It not only increased the amount of fraud we were detecting, but has reduced fraud cost by €37 million”.

Potential outcomes:

• With advance of AI in telecommunications, new models will be able to address a $39 billion dollar fraud problem
• In addition to preventing revenue loss, ongoing fraud detection efforts also boost brand reputation.

Robotic Process Automation

Robotic Process Automation (RPA) is revolutionizing the telecom industry by automating routine tasks such as data entry, order processing, and invoice management, merging the physical and digital worlds.

AI technologies are deeply integrated with these processes, making telecom operations more efficient. For example, AI helps telecom companies respond to custom demands and shorten the time it takes to bring new services to market.

Popular RPA tools include UiPath and Blue Prism, which offer unique features like drag-and-drop interfaces, process mapping, and cognitive capabilities that allow telecom companies to automate complex workflows, improving overall operational efficiency. These tools help companies meet modern challenges, ensuring faster, more accurate service delivery in an industry driven by increasing sensor data and network demands.

Potential outcomes:

• RPA can reduce operational expenses by 25% to 80% on tasks such as invoicing, data entry, and compliance monitoring.
• By automating back-office functions, employees can dedicate more time to strategic initiatives that require human insight, such as customer relationship management and innovation in service offerings.

Challenges and Solutions in AI Adoption

A recent Body of European Regulators for Electronic Communications (BEREC) report on AI adoption in EU telecommunications found that companies are facing significant challenges, particularly around issues of undetected data bias, liability in case of AI-related errors, and ensuring access to reliable data.

The AI-RAN Alliance also highlights key challenges for AI in telecom in their roadmap, such as AI hallucinations, high development and training costs for AI/ML models, and resource demands.

Below, we explore these challenges and practical solutions for AI adoption in telecom.

Managing Data Quality

The heterogeneous and non-standard nature of telecom data makes it difficult to train AI and achieve accurate predictions and insights. The larger the dataset is, the higher are the chances to find incomplete, inconsistent, or biased data points.

Inconsistent data quality can lead to suboptimal network performance and inefficient resource allocation.

To address these concerns telco experts implement data governance frameworks and standardization protocols to ensure data consistency. This includes automated data cleaning and normalization processes, using AI tools to prepare datasets for further analysis and storing it in data warehouses.

Explore top data warehousing tools for every use case in 2025

A report by McKinsey & Company also states that increased focus on data lifecycle is paramount for building a healthy AI-first infrastructure for telcos.

Overcoming AI Hallucination

At present, popular AI models don’t provide the required levels of reliability and safety to be trusted for critical infrastructure deployments.

Even though GPT-4 is the most accurate model, it still hallucinates 3% of the time. At this rate of making mistakes, deployment to network infrastructure will eventually lead to service disruptions and potential outages.

To ensure seamless integration and minimize risks of failure telecom operators must develop testing and validation procedures for AI models, incorporating techniques such as adversarial testing, redundancy, and failover mechanisms. It’s also crucial to establish clear guidelines for AI model deployment and monitoring to ensure reliability and safety.

Finding Technical Expertise

The fast pace of AI innovation in the telecom industry has led to a shortage of skilled professionals who can effectively develop, deploy, and manage AI technologies. Telecom companies often struggle to find professionals with the necessary AI and telecommunications infrastructure expertise. Another report by McKinsey found that by upskilling workers, telcos could reap enormous benefits. Some of the case studies cited in the survey include findings such as:

“One European telco recently increased conversion rates for marketing campaigns by 40 percent while reducing costs by using gen AI to personalize content.”

“A Latin American telco increased call center agent productivity by 25 percent and improved the quality of its customer experience by enhancing agent skills and knowledge with gen-AI-driven recommendations.”

These cases demonstrate how telecom firms must address the challenge of a lack of technical expertise by partnering with AI consulting firms or outsourcing specific projects to specialists. For instance, Deutsche Telekom has collaborated with AI startups to bring specialized knowledge to enhance customer service using AI. Upskilling existing staff through AI and machine learning training is also viable. Companies can thus cultivate in-house expertise while bridging the talent gap.

Future Directions and Innovations in AI for Telecom

Vodworks Helps Telecom Providers Build a Data Backbone for AI Integration

With substantial experience in telecommunications, Vodworks is an ideal partner for companies looking to implement AI solutions in the industry. The company has collaborated with leading telecoms, including True Digital, where it provided predictive analytics to track the spread of COVID-19, showcasing its ability to deliver data-driven insights for critical situations. With a diverse team of developers, data engineers, and solution architects, Vodworks efficiently bridges the gap between AI and telecommunications.

Reach out today to inquire whether we can help you with a personalized IT solution.