I still remember sitting in a darkened theater as a kid, watching Arnold Schwarzenegger’s Terminator materialize from the future. The idea of machines with superhuman strength and intelligence felt like pure science fiction — thrilling, but impossibly far away.
Decades later, I realize we didn’t get time-traveling cyborgs. Instead, we got something just as remarkable: machines that can predict the future, optimize decisions, and reshape entire industries.
AI is Unlikely to Fade Away
While artificial intelligence (AI) may still sound futuristic to some, its rise was almost inevitable. Labor costs rose while computing power became cheaper. Today, we have systems with massive computing power and problem-solving abilities that can take natural language input and return responses like a trusted colleague — but it wasn’t always this way.
AI has its foundation in statistical methods from the 1950s and has seen alternating periods of success and failure. After the rise of computer vision and digital personal assistants in the 2010s, AI started to transform everyday life. Now — with ChatGPT as the fastest-growing consumer application in history — it has found practical mass-market applications and is unlikely to fade.
Since AI systems recognize patterns in data and gain knowledge from it without being explicitly programmed, the use of AI reduces the need for human-made algorithms. For suitable use cases, AI can automatically create a model that identifies the best response to a wide variety of inputs, avoiding the need for writing traditional software. We can think of AI use cases within a quadrant: improving internal efficiencies on one axis and creating new market value through products on the other.
Like many other industries, telecommunications has widely adopted AI — and we expect it to lead to productivity increases and efficiencies. However, AI is only as effective as the quality of the data it relies on. This means data becomes the most valuable component — the fuel — of the whole AI system. Data must be stored in the right format, easily accessible, and ready for analysis.
The Perfect Storm That Changed Everything
In the 2010s, explosive data growth, increasing network complexity, and rising customer expectations created pressures that traditional reactive approaches could no longer handle. At the same time, subscribers became more tech-savvy. This shift opened new opportunities but also demanded some adjustments.
People were spending more time online than ever and expected their broadband to deliver — not just for leisure and entertainment, but also for productivity, content creation, and remote work. Consumers wanted instant access, real-time interactivity, and consistent performance, even at peak times. The tolerance for buffering and outages disappeared.
As an industry, we needed to track and identify trends and patterns, optimize resource planning, detect anomalies, and improve customer experience. These were areas where predictive AI techniques excelled and could surpass existing algorithmic power. More importantly, those first implementations created the data infrastructure, technical capabilities, and organizational expertise that would enable the next phase of AI evolution.
The business case for change was compelling. Anyone who has had to debug a failed network route or fiber cut knows that sifting through log files is a daunting and time-consuming task when trying to locate the problem. Network outages are very costly, leading to revenue loss and a poor customer experience. Tools such as traffic utilization analysis, anomaly detection, and predictive failure analysis, combined with increasing automation, allow operators to quickly recover from failures, and boost customer experience.
The Evolution of AI in Telecom
Many of the initial telecom use cases for AI focused on enhancing customer experience, reducing cost, and optimizing operations. Generative AI adoption has also grown within the telecom industry with the widespread use of chatbots. However, their results are often not satisfying in a business context. They can fail in solving critical problems and misalign with day-to-day operations. Despite the dystopian view of AI taking over the world or stealing our jobs, true transformation — with long-term productivity benefits and bottom-line financial impact — is rare.
Taking a step back and looking at the adoption of AI over the next decade, we expect the AI revolution to happen in three phases :
- First phase: Operators rely on predictive AI to improve the human decision-making process. AI is well-suited for analyzing lots of data and can predict what might happen next. It can recognize patterns, find trends, correlate data, and detect anomalies before they affect services. These expert tools can do the heavy-lifting and are trained for a specific job, giving highly accurate results. However, they require human oversight and cannot adapt to new situations.
- Second phase: An AI assistant is introduced, adding human-level reasoning capabilities. Instead of browsing a graphical user interface, you can interact with the AI assistant using natural conversational language. The assistant will self-learn to find the answer and — like a virtual engineer — consult different external resources on your behalf. You can get help with querying documentation, summarizing alarm information, creating a dashboard, or generating a NOC report. It’s not just a simple chatbot, but one that is connected to the cognitive fabric of the network.
- Third phase: Agentic AI is implemented, taking humans out of the loop. Where AI assistants wait on human prompts for every action, AI agents operate autonomously to plan and pursue a goal (e.g., an AI agent can detect an anomaly, find the root cause, reboot an ONT, and proactively notify subscribers without human intervention). It can also open a ticket in the field service tool to dispatch an engineer and verify if the problem is resolved after the intervention. At this point, different agents will work together and learn how to solve an E2E network problem collectively. This is the power of agentic AI — but we’re not there yet.
The Foundation Is Set
The industry adoption of AI is just beginning. Predictive AI’s success has helped advance several critical foundations in data infrastructure, intent-based interfaces, and operational expertise. These elements will prove essential as the industry moves toward generative and agentic AI applications. It won’t just change how networks run. They’ll redefine how services are designed, delivered, and experienced, changing the telecommunications industry forever.
David Eckard, Head of Strategy, Fixed Networks
Nokia
David Eckard serves as Nokia’s Head of Strategy for Fixed Networks. Since 1999, he has been a key advocate and industry leader in the creation and adoption of fiber access in the U.S. and globally. Bringing a multi-disciplinary perspective, David advises executives, policymakers, and industry leaders on market and technology challenges. He has held leadership roles including VP of Strategy and Technology for North America, VP of Business Strategy for Optical Networking, and CTO of Fixed Networks. He has guided the development of key technologies that make Nokia a trusted partner for next-generation networks.
