After years of intense investment cycles in 3G, 4G, and now 5G, we find ourselves at an inflection point. Revenues are plateauing, operational costs are escalating, and the long-promised profitability boost from new networks has yet to fully materialize. Against this backdrop, you may have seen AI heralded as a silver bullet. A catch-all solution.

In a year where Agentic AI is pushing the boundaries of what’s possible, it's easy to think full automation is inevitable. But it’s worth considering where humans sit in all of it.

We’ve guided telcos through industry and technological shifts – from analogue to digital networks, hardware to software -defined systems, and the rise of cloud-native infrastructure. We’ve seen firsthand how the prevailing narrative around wholesale automation is deeply flawed. You can’t deny the power of Large Language Models (LLMs). But in such a complex and regulated environment, it falls on us to shoulder the finer details of operational control.

The future lies in "micro-augmentation" AI that works tirelessly at the service of human expertise and judgment. It's a smarter path to transformation. One that respects the industry’s intricate knowledge base while still delivering on the bold promises that AI has made.

The limits of full-scale automation

The idea of end-to-end AI-driven operations is hard to resist, but the reality is far messier. Even as autonomous agents become more capable, real-world telco environments are still convoluted, heterogeneous, and highly regulated. As an industry, we operate vast networks that span decades-old infrastructure alongside cutting-edge 5G technology.

Managing these networks demands a deep contextual knowledge and the ability to make quick decisions under a huge amount of uncertainty. That’s where automation begins to fray. Add a whole host of convoluted regulatory requirements, and you’ll find that full-scale automation isn’t up to the job.

Despite remarkable progress, LLMs and other AI models still struggle with explainability and biased data. A 2024 McKinsey Global AI survey found that 40 percent of respondents identified explainability as a key risk in adopting generative AI systems; yet only 17 percent were actively working to mitigate it. This tells us that trust and transparency are being lost in the pursuit of greater efficiency. Without reliable, transparent decision-making, going into automation blind is a recipe for disaster. From network outages to compliance breaches, you’re looking at a level of reputational damage that is hard to undo.

Then there’s the fact that much of our operational knowledge is tacit, accumulated through years of hands-on experience. Replacing skilled professionals with opaque algorithms threatens to erode this valuable knowledge base. Without it, you risk creating brittle systems that will ultimately crumble under pressure.

Micro-augmentation: A smarter strategy

Micro-augmentation offers a more human-centered way forward. Rather than chasing full automation, it focuses on using AI to enhance specific tasks and workflows incrementally while keeping humans at the heart of change. This means deploying AI where it can offer the most tangible benefits.

Take network design, for instance. It’s a process that’s long been bogged down by time-consuming documentation and repetitive input. We’re now seeing AI Agents step in to support solution architects by semi-automating the drafting of High-Level Design (HLD) documents.

Instead of starting from a blank page each time, engineers are leaning on AI to take care of the heavy lifting. Instead of getting bogged down in manual work, engineers can channel their energy into the nuanced decisions that will ultimately shape the future of network design. With AI taking on the grunt work, they’re free to make the intuitive judgment calls that still need a human touch.

We’re also seeing some really promising work around surfacing institutional knowledge. This is the kind of knowledge that often lives in the minds of a few experienced individuals or is buried deep within siloed and scattered company systems. By building AI-powered repositories in-house using generative AI agents, telcos are making that collective wisdom far easier to tap into.

Beyond the obvious time-saving advantages, it’s also a great example of how micro-augmentation can encourage greater collaboration and cross-functional upskilling.

Because AI excels at automating repetitive, rules-based tasks, it's proving invaluable for large, resource-heavy transformations. This could be rolling out a new network or decommissioning legacy infrastructure. These are the sorts of projects that typically demand huge amounts of time and manpower, but with AI agents, telcos can bring their costs down and shrink their timelines without sacrificing quality.

These are just a few examples of how micro-augmentation can drive incremental improvements that add up. By enhancing how people work, rather than replacing them, it preserves the institutional knowledge and problem-solving capabilities that are critical to business resilience.

Why headcount reduction is a dead-end

There is a temptation to frame AI-driven transformation around workforce reductions, particularly as the industry grapples with tight margins. But the reality is that blunt headcount cuts are a superficial fix. They may deliver cost savings in the short-term, but ultimately undermine long-term resilience.

Research shows that organizations capturing the greatest value from AI are those applying it to core business functions like product and service development, rather than focusing solely on workforce reductions. This tells us that talent is still a key differentiator in telecoms – and will continue to be as the industry shifts towards more software-driven, services-oriented models.

Rather than chasing automation purely to eliminate roles, prioritize investing in retraining and reskilling your workforce to thrive alongside AI. This could mean a variety of things - upskilling engineers to work alongside predictive maintenance tools, training customer service agents to use AI copilots during interactions, or even developing new hybrid roles that blend technical expertise with human insight. Whatever the approach, the bottom line is that humans bring something to the table that machines fundamentally cannot.

A call to pragmatism

We’re under immense pressure to deliver better margins, faster services, and more resilient networks. There’s no doubt that AI will be at the heart of this next chapter. But the path to success will not be paved by unrealistic expectations of automation.

Instead, it’s important to embrace an incremental approach. This means deploying AI to amplify human strengths. Even as we explore the use cases for Agentic AI, micro-augmentation remains the best way to unlock the full potential of AI while safeguarding the operational excellence that telcos have spent decades building.