Telco AI planning options

April 1, 2026

There's been lots of buzz going around the telecom industry about AI. If you're wondering how to add AI to your telecom business, there's a study available with lots of information on that. Let's have a look.

You've probably heard about, and maybe used, AI chatbots like ChatGPT, Google Gemini, Claude, and Copilot. Is this what other telcos are using for AI? We can't speak for all of them, but we don't think so.

There's a study, Large Language Model (LLM) for Telecommunications: A Comprehensive Survey on Principles, Key Techniques, and Opportunities, that explains how one could go about putting AI in a telco. It's a lengthy, in-depth article. We'll quickly summarize the highlights here. We think this will help you understand the most likely options to put AI in a telco.

You'll see the term “LLM” all over this material. It stands for “Large Language Model.” It's the software and data behind the AI chatbots that users interact with when they're using AI.

Where do you get an LLM that knows telecom?

Telecom domain-specific information is rare in general-domain LLMs.
- Applying a general domain LLM to telecom tasks may lead to poor performance.
For telecom applications, pre-training an LLM from scratch can be time-consuming.
A more efficient approach is to fine-tune a general domain LLM for specific telecom tasks.
- Open source LLMs are available. These can be augmented and trained for telecom.
- Not a small task, but much easier than building a telecom-specific LLM from scratch.

You can't buy telecom AI resources off the shelf, but you can take an open source LLM and add telecom knowledge to it.

Applications in telecom

Question answering, user feedback processing, literature review, and troubleshooting with telecom domain knowledge.
Code generation, including refactoring to improve the efficiency and reliability of code.
Complex project management with multi-step scheduling.
Network configuration support, including security and compliance configuration, fault diagnosis, and troubleshooting to improve network management.
Detection of cyber-attacks, contributing to mitigation and recovery strategies.
Traffic analysis

You might be thinking, “Wait a minute, I'd never let AI do some of these things. We use engineers who know what they're doing.”

You're correct to be cautious. Think of this as providing tools to help your staff do their work more effectively and efficiently. You can also provide AI for customer service staff. Once you're comfortable and feel that it's accurate and usable, then you could also make it available to your customers.

Deployment strategies

Central cloud. Pro: substantial resources. Con: long response time, high bandwidth costs
Network edge. Pro: shorter response time, less bandwidth. Con: less capacity.
On device. Pro: fast, real-time service. Con: small models must be well-trained.
Cache. Store full precision parameters in the central cloud, quantized parameters in the network edge, and frozen parameters in the user device. Pro: reduced latency. Con: complicated.
Cooperative. Large model in the central cloud periodically updates small models. This is a variation on the Cache method and has similar pros and cons.

The interesting thing about these strategies is that they break down the myth that AI can only be run in the cloud. Yes, the largest, most powerful models require massive resources. But there have been advances in smaller models, and they can produce good results in some use cases. They're catching up.

More information

The full study is available online: Large Language Model (LLM) for Telecommunications: A Comprehensive Survey on Principles, Key Techniques, and Opportunities, by Hao Zhou et al, 2024, IEEE Communications Surveys and Tutorials.

It's a long, technical paper, but there's lots of good information in it.

We also reviewed some recently announced AI plans and summarized what they're doing and how it works in a recent blog post, Artificial Intelligence deployment in telecom.

TransNexus solutions

TransNexus is a leader in developing innovative software to manage and protect telecommunications networks worldwide. The company has over 25 years of experience in providing telecom software solutions including toll fraud prevention, robocall mitigation and prevention, CDR and call analytics, advanced call routing, billing support, STIR/SHAKEN, and branded calling.

This information will only be used to respond to your inquiry. TransNexus will not share your data with any third parties. We will retain your information for as long as needed to retain a record of your inquiry. For more information about how we use personal data, please see our privacy statement.