The Telecom AI Paradox: Big Revenue Hopes Meet Agentic AI Skepticism

The Agentic AI Readiness Gap: Why 57.5% Say "Not Yet"

If revenue optimism is the headline, the subheading is a firm yellow light on agentic AI. The survey found that 57.5% of respondents believe the complexities of telecom networks have not been properly baked into agentic AI developments . The panel at the DSP Leaders World Forum, which included executives from Telefónica, Wind River, and Appledore Research, validated these concerns with on-the-ground technical arguments .

Agentic AI refers to systems that can plan, execute multi-step tasks, and interact with other agents autonomously—far beyond current AI assistants that respond to single prompts. For a telecom production network, an agentic system might independently reroute traffic, spin up virtualized network functions, or negotiate service-level agreements with another operator’s agent in real time. This sounds powerful, but it introduces entirely new failure modes in infrastructure where human lives and critical services are at stake.

The Protocol Problem: Why Agent-to-Agent Communication Isn't There Yet

The technical linchpin for multi-agent systems is the communication protocol. Two emerging standards—the Model Context Protocol (MCP) and Agent-to-Agent (A2A) protocols—are central to the vision of interoperable AI agents. But the survey reveals only 30% of respondents believe understanding and using these protocols is a game changer for telcos today .

Panelists at the DSP Leaders World Forum deepened this critique. They pointed out that these protocols are extremely young—formulated only a couple of years ago at most—and their real-world deployment has been limited to very closed, single-vendor environments . For a heterogeneous telecom network running equipment from multiple vendors across different generations of technology, this lack of open, proven interoperability is not just a maturity problem—it is a fundamental architectural gap .

A panel of experts discussing the findings noted that the main barrier, beyond the maturity of the AI models themselves, is that agent-to-agent communication currently has no credible path to work across multi-vendor, production-grade telecom infrastructure. Until protocols are tested, standardized, and shown to be secure in open environments, deploying agentic AI at scale remains a bet most operators are not willing to make.

Trust and Sovereignty: The Twin Pillars of Telco AI Adoption

Running through both the survey and the forum discussions is a dual theme: trust and digital sovereignty. Trust is the broader, less technical barrier. Telecom operators are accountable for network uptime, data security, and regulatory compliance. Handing decision-making authority to AI agents requires a level of confidence that the current technology cannot yet provide .

The sovereignty conversation adds a geopolitical and commercial layer. The survey found that 54% of respondents view sovereign AI as a strong business opportunity for telcos. Another 27% believe it should be left to IT specialists, while 19% remain unsure .

Sovereign AI refers to AI systems and infrastructure that are designed, built, and operated within a specific country or region, under local laws and data governance frameworks. For enterprises and governments that cannot risk their data flowing through foreign-controlled cloud services, telcos are uniquely positioned: they already operate trusted, regulated national infrastructure, control data center real estate, and have deep customer relationships. As one panel discussion noted, telcos are in a very good place to be the right partner for enterprises that need sovereign AI guarantees .

Edge AI: Where the Network Meets the Data

The edge computing layer—where data is processed close to its source rather than in centralized clouds—is where sovereignty, trust, and AI converge on the network itself. Forum discussions highlighted that network edge struggles are directly tied to AI and trust dynamics. As AI workloads increasingly require low latency and data localization, the network edge becomes the natural enforcement point for sovereignty policies .

The challenge is cost. Data sovereignty requirements add expense: specialized hardware, compliance overhead, and the complexity of maintaining distributed compute across thousands of edge locations. Telecom operators are grappling with how to price and package edge AI services when the underlying sovereignty costs remain difficult to quantify and pass through to customers .

A Split-Screen Future for Telecom AI

The emerging picture is not a simple story of adoption or resistance. It is a split-screen view of AI in telecom: the industry is enthusiastically pursuing AI for revenue growth and new services, while simultaneously pumping the brakes on the most autonomous form of AI that might one day run the networks themselves.

Current AI applications in telecom are focused on anomaly detection, customer interaction, and operational support—areas where human oversight remains the final safeguard . The leap to agentic AI, where systems act independently across vendors and network layers, is where trust, protocols, and sovereignty concerns converge into a barrier that 57.5% of the industry is not ready to cross today .

As the DSP Leaders World Forum discussions made clear, the industry is not rejecting agentic AI—it is demanding that the underlying protocols, interoperability frameworks, and trust mechanisms mature before production deployment becomes realistic . In the meantime, sovereign AI and edge services represent a more near-term, trust-based business opportunity that leverages telcos' existing infrastructure advantages.