Clearing the legacy problem 

The drive for productivity, in both telecoms and their customers, is not a new paradigm. What is new is that the latest generation of AI-assisted development tools has given it genuine potential. Developers are seeing 45% average productivity gains in designing, implementing, and managing new code, along with the ability to autogenerate documentation, optimization, modernization, and the securing of existing code that may have been written 30 years ago.

This kind of tool is built to take the legacy problem head-on. For modernization of older code, it reads back through that code, creates documentation, creates architecture diagrams, and reasons across what those programs are actually doing. It looks at the code from a governance and security perspective, because everything about the approach is premised on governance and security. It suggests what to modernize, what to keep, what to translate into a different language, and where parallel programming could replace sequential and lift performance. It is not an LLM for programming. It is a partner for development. For new code, that shows up as design assistance, architecture diagrams, code, test cases, and testing, all of which can be auto-generated with human guidance. What is notable about the better tools is multi-model orchestration, automatically routing each task to a suitable model based on accuracy, performance, and cost. They can automate full software development lifecycle workflows while governance, compliance, and security controls are built into every step.

This changes the math on how quickly an operator can credibly offer the agentic services its enterprise customers want.

The tailwinds ahead 

Two things on the horizon are worth naming.

The first is sovereignty. The regulatory direction across most of the world is moving toward stronger requirements for not only where data lives, but who can access it, what operational control can be demonstrated, and how compliance is proven. Telcos are natural partners to the government. They are used to regulation, they have the scale and the skill set to build complex platforms, and they have the kind of customer relationships that hyperscalers do not. The sovereignty conversation over the next few years is going to favor the operators who have already invested in the underlying architecture. It is going to be uncomfortable for businesses still running on someone else's hyperscaler with limited control over the stack.

The second is quantum. We are doing work now with customers, including in telco, on use cases that lean on the kinds of mathematics and probability that quantum handles very well, and that classical systems cannot. Materials science is one of them. Better batteries that last more than a day on a phone would change consumer expectations for an entire industry. We can investigate network optimization problems with exponential variables while exploring countless end states at once. There is more coming in the space between AI and quantum than most people realize.

Where this lands 

I started with the GSMA finding because it inverts the usual framing. The conversation about AI in telecom is too often a conversation about catching up about the cost of falling behind hyperscalers, about the disadvantage of legacy infrastructure, and about the gap between what enterprise customers want and what CSPs can offer.

The data does not support that framing. Enterprise customers, especially the 90% of the market that SMEs represent, are looking at telcos as the partner they trust most. They are waiting for the offering. The operators that get there first, with a sovereign platform underneath, with agentic services on top, and with the legacy work done so the conversation is about value rather than constraints, are going to define the next decade of enterprise revenue in this industry.

The concern about missing out on AI is real. The concern about messing it up is just as real. The answer is not to pick one. The answer is to move now with partners who have done the work to make AI consumable, usable, and trustworthy, and to use the relationship you already have with your customers to build something they cannot easily get anywhere else.