The future of AI skills in telecoms: Fireside chat with LabLabee

Download

Artificial intelligence (AI) has gained significant momentum in the last couple of years and continues to reshape the telecoms industry. While the promise of autonomous networks and intelligent service operations is clear, many operators are still navigating how to integrate AI effectively given the rapid pace of evolution and its relative nascency. STL Partners sat down with LabLabee, a lab-based training platform for telco engineers and developers, to discuss where AI is making an impact today, what challenges remain, and how telcos can upskill their workforce to take advantage of this technology.

Where do you see AI being deployed within telcos, and what do you think its impact will be?

LabLabee: AI is a very trendy topic in telecoms, but the level of implementation varies widely. Most deployments today are small-scale pilots and use case maturity is limited to a few key areas. This is because AI isn’t a simple pass – it’s a long and complex process that involves many moving parts. Telcos envision big opportunities in using AI to improve agility and optimise cost structures.

We’re already seeing a range of live, customer facing deployments today. AI can assist with customer interactions, supporting salespeople to better understand customer needs and offer dynamic product and pricing options for core products and services. I’ve also seen examples of AI used in contact centres when training new sales hires. For these use cases, AI has already demonstrated proven results.

However, the major gain AI offers telcos is network automation, where AI can detect issues faster and minimise costs. The complexity arises from the need to match multiple technologies and types of data, and that AI has an understanding of how different elements and network layers work together. In a recent conversation with a Tier-1 Middle Eastern operator, they explained that often network failures happen when multiple small issues occur at once, and a single alert is sometimes meaningless without the wider context of network issues in which it occurs. AI can help correlate the different signals and filter out unnecessary noise to identify the root cause of a problem.

However, truly autonomous networks – where AI can understand intent and independently interact across all the network and technological layers (infrastructure, hardware, software, security, etc.) – are still several years away. The biggest value will come from automating core network infrastructure and services. The journey is complex because networks evolve constantly, there are several dependencies, and AI requires vast amounts of high-quality data with accurate documentation to develop a clear view of what is good and not good, and to train, test and validate AI systems are effectively delivering intent.

What are some of the biggest challenges preventing telcos from scaling AI?

LabLabee: In AI, what we’ve heard is that there are a lot of amazing things it can do. It works very well in a limited scope or under specific conditions, but there is still some work to do before AI can be scaled and we see it fully operational in production. Telcos must start by identifying the challenges which limit AI from delivering progressive value. The main challenges come down to skills, integration

Some operators simply don’t have the right internal AI expertise, time or resources to move fast. Others have data scientists and software developers who don’t fully understand the different business-related functions, core behaviours and network constraints of telecom operations, which means problems often appear at the testing stage. Some network engineers, on the other hand, know the business side very well, but don’t understand how AI will be most relevant and how far it can go. This slows progress and creates siloes.

Telcos need to be more agile and focus on simplifying how teams’ experiment. Accessing testing environments and tools becomes key to avoid delays and ensure smooth and impactful deployments. To overcome this, many telcos are creating cross-functional centres of excellence that bring together AI, software, and network experts to work collaboratively. This helps align technical and business priorities while encouraging knowledge sharing.

With integration, one concept that’s helping move things forward is the MCP model, introduced by Anthropic. This is an open standard for connecting AI assistants to external systems and data stores. It acts like an API for AI, allowing systems to talk to each other. Each service can have its own MCP server, and in an ideal world, AI could connect to all these servers across different business and technology domains to discover data, context and features automatically. But this only works if people build and train those systems well.

For telcos, this is complicated because they have many dimensions and layers all with different variables. To reach the real value proposition of autonomous networks, AI needs to be rechecked every time something changes to make sure it’s performing correctly. In tandem, other domains of the network are also evolving. Any tech changes can have strong impact on new developments, and model training and testing are necessary to ensure AI will work appropriately. To summarise, strong AI models, telco expertise, good data, and agile testing capabilities are among the key elements to scale AI.

How can telcos build the right skills for AI internally?

LabLabee: The key is to make learning continuous, practical, and cross-functional. The first step is onboarding AI teams who really understand how telecom networks and operations work in practice. Many AI experts come from software backgrounds, so giving them hands-on experience helps them apply their work to real network operations. This eases interactions and collaboration between AI experts and network engineers and allow teams to move from conceptual knowledge to practical experience.

From an organisation point of view, we also believe creating small, agile teams and centres of excellence where you mix different skills and competencies is important. This approach creates opportunities to transfer knowledge between AI specialists, network engineers and business experts, solving the problem of organisational siloes and building next generation experts. In the long term, telcos need to focus on both technical upskilling and change management to ensure teams trust and understand how AI systems make decisions.

We’ve seen this work well at a major European operator, where engineers are encouraged to spend several weeks working in different teams. It helps them understand each other’s jobs, share different methods of working and, over time, build confidence in testing new ideas, tools and processes.

What role can AI play in training and upskilling telecoms engineers?

LabLabee: We see huge potential for AI-driven learning. Ultimately, it enables a new way to package and deliver training that’s flexible, responsive, and better suited to the fast-changing telecoms environment – in essence, allowing engineers to develop a set of skills which are complete and enables them to be fully operational. Instead of traditional, static learning, AI can make training far more dynamic and productive.

We often talk about the use of AI Copilots to support engineers on specific tasks and contextual-based problems – like a role-based companion for developing automation or even AI itself – enabling engineers to learn directly as they work. These companions can share relevant knowledge in real time, help identify skill gaps, and suggest tailored training plans based on the challenges an engineer faces in the field.

For example, an AI assistant could observe how an engineer troubleshoots a network issue and provides feedback or suggest additional learning modules. This creates a learning loop where engineers get immediate support to solve a problem and receive additional training to strengthen their long-term skills. It’s a very different approach from standard training sessions which takes place annually – the learning happens right when it’s needed.

At LabLabee, we’ve seen strong results when operators embed training into real-world projects, allowing engineers to learn by doing. Instead of teaching large blocks of theory, we create targeted, hands-on labs that focus on specific real-world scenarios, like configuring permissions or enabling a new feature. This kind of learning reduces handovers, speeds up problem-solving, and ensures engineers actually gain the skills themselves rather than letting AI do the work for them.

If you’re interested to learn more about LabLabee, you can visit their website at https://www.lablabee.com/.