Artificial Intelligence (AI) has ascended to the very top of the boardroom agenda for telecommunications providers globally. As operators race to capture the efficiencies promised by machine learning and generative AI, the industry faces a paradoxical challenge: while demand for AI-driven transformation has never been higher, the structural complexity of telco networks remains a formidable barrier.
To bridge the gap between experimental pilots and enterprise-wide deployment, the GSMA and TM Forum have announced a strategic, industry-wide collaboration. By aligning the GSMA’s focus on "telco-grade" model development with TM Forum’s robust architectural frameworks, the two organizations aim to move the needle from fragmented, isolated innovation to a cohesive, standardized AI ecosystem.
The Core Challenge: Why Telecom is Not "Just Another Industry"
For many sectors—such as retail or finance—AI adoption has been relatively straightforward, leveraging generic frontier models to enhance customer-facing services. Telecom, however, operates in a different reality. Modern networks are multi-vendor environments, characterized by highly fragmented data silos and an almost zero-tolerance threshold for service disruption.
Louis Powell of the GSMA and Andy Tiller of the TM Forum argue that the current generation of AI models is simply not equipped for the unique nuances of network operations. These models are typically trained on general-purpose internet data, leaving them ill-equipped to understand the complex language, architecture, and operational logic inherent in telecom infrastructure.
The Network Layer Paradox
The current state of AI in telecom is heavily skewed toward customer experience (CX) and enterprise support functions. GSMA Intelligence data reveals a striking disparity: while networks account for roughly 34% of total operational expenditure (OPEX), only 16% of AI deployments are actually targeting the network layer. This misalignment stems from a lack of "telco-native" intelligence. To effectively automate network maintenance, load balancing, or predictive maintenance, AI must be highly accurate, efficient, and, above all, trusted from the moment of implementation.
Chronology of the Collaborative Shift
The path to this collaboration was paved by a series of industry-wide realizations regarding the failure of "off-the-shelf" AI solutions.
- Early 2023: Recognition of the "AI Pilot Trap." Many operators reported that while Proof-of-Concept (PoC) projects were successful in sandboxes, they failed to scale due to integration hurdles and lack of interoperability.
- Late 2023: The GSMA launches the Open Telco AI initiative, focusing on the creation of specialized, telco-grade models (such as the OTEL family) trained on proprietary telecom datasets.
- Early 2024: TM Forum intensifies its focus on the Open Digital Architecture (ODA), emphasizing the need for standardizing how AI agents communicate across multi-vendor networks.
- Mid-2024: The formalization of the GSMA-TM Forum partnership. Both organizations recognize that models (GSMA) without frameworks (TM Forum) remain academic, and frameworks without models remain empty shells.
- Present Day: Initial deployments, such as those at Globe Telecom and AT&T, serve as the first "live" test cases for this unified approach, proving that the theory of integrated AI can be converted into tangible operational gains.
Supporting Data: The Case for Standardization
The economic argument for this collaboration is rooted in the urgent need to manage rising OPEX. As networks become more sophisticated—incorporating 5G Advanced and preparing for 6G—the cost of human-led management becomes unsustainable.
- 34%: The proportion of total operator OPEX currently tied to network operations, representing the largest area of potential AI-driven cost reduction.
- 16%: The current share of AI investment directed toward the network, suggesting a massive untapped opportunity for efficiency gains.
- Multi-Vendor Friction: Research indicates that the primary cause of failed AI scaling is the inability of models to interact with diverse vendor APIs, a challenge TM Forum’s ODA is designed to solve.
Official Responses and Strategic Vision
The collaboration is not merely an administrative alignment; it represents a fundamental shift in how the industry views "telco-grade" AI.
The GSMA Perspective
Through Open Telco AI, the GSMA is tackling the "intelligence" deficit. By creating shared foundations of models trained on specialized telecom data, the GSMA aims to ensure that AI can interpret network telemetry as effectively as a seasoned network engineer. "We are building models that don’t just guess, but understand the structural constraints of a telco environment," note representatives from the GSMA.
The TM Forum Perspective
TM Forum brings the "scaffolding" required to house this intelligence. Through its Autonomous Networks and Trustworthy AI missions, the organization is defining the "rules of the road." Andy Tiller emphasizes that for AI to be useful, it must be interoperable. "We are providing the frameworks, standards, and solution packs that allow an AI model—regardless of who built it—to communicate across a multi-vendor, cloud-native architecture," Tiller explained.
Real-World Impact: From Theory to Deployment
The effectiveness of this partnership is already being tested in the field, moving beyond industry white papers into the live network environments of major global operators.
Globe Telecom’s Multi-Vendor Breakthrough
In the Philippines, Globe Telecom is demonstrating the power of this unified approach. By utilizing TM Forum’s technical solution packs, Globe has been able to normalize APIs across different RAN (Radio Access Network) vendors. Simultaneously, by applying Open Telco AI’s specialized models, the operator is automating root cause analysis, significantly reducing the time required to diagnose network faults.
The AT&T Integration
AT&T’s work with the OTEL family of models is another benchmark for the industry. By adapting these specialized telco models into TM Forum’s Model-as-a-Service (MODaaS) and ODA Canvas, AT&T is effectively turning proprietary innovation into a standardized, interoperable asset. This allows other industry players to adopt proven, operator-grade AI solutions rather than reinventing the wheel in isolation.
Implications: The Road to the AI-Native Network
The implications of this collaboration extend far beyond immediate cost-cutting. By aligning on shared assets and common standards, the telecom industry is positioning itself to become a foundational pillar of the broader AI economy.
1. Accelerating the Move to Autonomous Networks
The ultimate goal is the "AI-native" network—an infrastructure that can self-heal, self-optimize, and self-configure. The collaboration between the GSMA and TM Forum provides the roadmap for this evolution, ensuring that the transition is collaborative rather than chaotic.
2. Reducing Vendor Lock-in
By standardizing the way AI interacts with network components, operators gain the flexibility to choose the best models and the best network hardware independently. This reduces the risk of being tethered to a single vendor’s proprietary AI stack, fostering a more competitive and innovative marketplace.
3. Creating a Trusted Ecosystem
AI in telecom requires high levels of trust. By developing standardized benchmarks and evaluation criteria, the industry can ensure that AI agents acting on network traffic meet rigorous safety and performance standards. This is essential for the transition to critical services, such as industrial IoT and autonomous driving, which will rely on the reliability of telco networks.
Future Outlook: Looking Ahead
As the industry gathers for key events like DTW Ignite in Copenhagen and MWC Shanghai, the message is clear: the era of fragmented AI development is ending. The focus is shifting toward "end-to-end alignment"—where use cases, models, and deployment pathways are designed as a unified whole.
The coming year will be critical. The GSMA and TM Forum have committed to releasing further data, model evaluations, and member-driven proofs of concept that will continue to codify what "telco-grade" actually looks like. For operators, vendors, and technology partners, the directive is to stop navigating disconnected efforts and start contributing to the shared foundation.
Telecom sits at the heart of the global AI economy. By finally aligning its own house—through common standards, shared datasets, and collaborative frameworks—the industry is poised to move from being an enabler of other people’s AI to becoming an AI powerhouse in its own right. The tools are being built, the frameworks are being set, and for those ready to align their ambition with the industry’s shared vision, the opportunity is transformative.
