Telecommunications, along with every other industry, has fully embraced the promise of AI from streamlining operations, to enhancing customer service experiences. According to NVIDIA’s State of AI in Telecommunications Report, 60% say their organization is using or assessing generative AI and 89% have a plan to increase AI spending in 2026, up from 65% a year ago. While this experimentation is widespread, scaling these efforts into real-world impact remains a challenge. Operators are quickly discovering that what might work in a controlled environment won’t easily translate into complex, live production systems.
Executives across the industry are all beginning to shine a light on the beginning of a shift from innovation to execution. Even with this momentum, though, many operators are finding themselves stuck in the “pilot purgatory” phase, where promising initiatives demonstrate real value, but fail to make the leap into full-scale deployment.
The next phase of AI in telecom will not be defined by new ideas or next-level innovation alone. It will be defined by the ability to operationalize AI at scale across systems, workflows and customer interactions, rather than treating it as a top layer.
The Scaling Challenge: Why AI Stalls After the Pilot Stage
Despite heavy investments and experimentation, recent data shows that most communications service providers (CSPs) are still early in their AI maturity, with 84% being at the co-pilot/preliminary stage – far from realizing AI’s full potential. This shows that the true challenge is no longer proving that AI works. There are already countless examples of successful pilots and promising use cases to show for it.
The real problem is making it work within the realities of today’s telecom environments. AI models may perform well in isolation, but scaling them across live, interconnected systems is way more complex. Operators must now work with legacy infrastructure, fragmented data systems, and the challenge of integrating AI into existing decision-making processes.
The execution of AI in telecom is also more difficult given the nature of the industry being a highly regulated, always on environment where even the smallest disruption can have significant consequences. This all adds an extra layer of technical and operational complexities and contributes heavily to where these initiatives tend to stall.
Rising Expectations Are Outpacing Execution
At the same time, we’re seeing expectations around AI accelerating rapidly because it’s embedded in so many facets of our lives. Within the next two years, operators expect the majority of customer interactions to become AI-driven, which is a major sign that the way services are delivered are continuing to evolve. Consumers are already showing they’re open to this transition with 77% reporting a baseline level of trust in AI agents, and 69% saying that effective AI interactions positively influence their perception of a brand, according to Amdocs data.
However, this growing acceptance is accompanied by equally high expectations. AI is no longer being evaluated based on novelty alone, it’s being judged on its predictability across operating systems and performance through customer-facing channels. This means that users want interactions to be seamless, context-aware and capable of delivering reliable resolutions. They also expect continuity across channels and the ability to escalate their issues to a human when needed and without friction.
As a result, we’re seeing the gap widen between the expectations of AI, and what telecom operators are providing at scale. This gap, though, is no longer purely a technical issue. It’s a brand and trust issue – and every single interaction with AI now has the potential to strengthen, or weaken, customer relationships, meaning getting it right isn’t an option.
What It Takes to Operationalize AI
The focus should now be closing this gap, but this will require embedding intelligence into how telecom already runs – moving from isolated experimentation to fully integrated, operational systems.
Operators should focus on four key areas:
- Integration: This will all start with integration into the core systems that run telecom – specifically OSS and BSS environments that power billing, customer management, and network operations. Operators must ensure that AI can be embedded into these existing systems without requiring costly and disruptive rebuilds. The ability to connect AI into already complex environments where data spans network elements, operational systems and customer platforms will be essential for scaling the technology.
- Orchestration: It also requires orchestration across domains that have traditionally operated in silos. AI must be able to coordinate across customer care, billing, and network operations simultaneously. For example, resolving a customer issue may involve combining network diagnostics, service history and billing context in real time. This is what will allow AI to move from isolated use cases to end-to-end journey support across OSS, BSS and customer engagement layers.
- Governance: Equally as important, ensuring transparency, compliance, and control will be critical as AI takes on a more active role in decision-making. This is especially important in telecom given regulatory requirements and the real-world impact of service disruptions.
- Measurement: Finally, many AI initiatives in telecom fail to scale because they lack clear, measurable links to business outcomes. Proving ROI – whether through reductions in average handling time (AHT), higher resolution rates, shorter mean time to repair (MTTR), or increased revenue – will be critical to moving from experimentation to execution at scale.
Together, these elements will help AI evolve from a standalone capability into a coordinated, accountable system embedded directly into telecom operations.
The Next Phase: Execution at Scale
Telecom is no doubt entering a defining phase of AI transformation, but this will take hard work and dedication from the entire industry. The need is no longer for more pilots. We now need operational AI that delivers consistent, real-world outcomes at scale – AI that is built into the systems that power network operations, customer care, and revenue management.
The organizations that see success will be the ones who move beyond experimentation and testing stages, and focus on embedding AI into core business processes, aligning the technology with operational and organizational readiness. This means improving efficiencies across operations, elevating customer interactions through faster and seamless service and unlocking measurable revenue outcomes.
AI has proven its value in pilots, that’s no longer a question. Now, we need to answer if it can deliver value where it matters the most, in live environments, at scale, reliably and in ways that build strong trust with customers.
