The Long and Winding Road to Agentic AI for Telcos

Recently, we had the opportunity to attend, absorb and learn from a nearly breathless two days with Amdocs at an event for analysts based across the wider APAC region. Engagement with a market leader that operates globally and has such a wide portfolio of products and services can leave one with cognitive overload. However, this event threw up important themes and a lot of food for thought.

The telco industry is changing, and Amdocs is too

The monolithic mobile operator network with a primary focus on building macro coverage has been the predominant model for decades. This model has been shifting for some time and there is increasing evidence of new service provider models emerging. Amdocs CMO Gil Rosen laid out these shifts in a keynote address and laid out the case for this shift in some detail. Rosen put forth a number of new telco “archetypes” which are throwing up new service provider models, covering the traditional near ‘utility’ type providers to emerging models focused specifically on consumer (MVNOs and digital lifestyle “TechCos”), enterprise (MVNE), NetCos and others.  

Much of this vision is already in evidence though the degrees vary by geography and market/operator maturity. What this does imply is that there are new opportunities emerging for vendors like Amdocs, and the company is moving in that direction, with a new strategy that will see them expand their offerings across domains and service provider types. Amdocs has already spread its wings beyond its core mobile operator customer base, with wins in Media and fixed line.

While many telcos will focus on adopting strategies that will make them leaner, more efficient and able to connect ‘anyone, anywhere’, there are a select few large global and regional players who are already thinking beyond, investing in and positioning themselves to become true “TechCos”. Rosen laid out a vision for how these providers will evolve beyond their traditional methods of market and customer segmentation towards “personality engineering” by using AI-powered personal agents to offer highly personalized and proactive customer experiences. This shift requires sophisticated AI agents to be deployed within the agent. In effect, transforming the traditional telco into an “AgentCo”.

Can the telco lead forward, beyond Digital to a Cognitive Organization?

Digital transformation has been a mission statement for most operators over the last few years, with a focus on moving to cloud-native infrastructure, investing in customer experience platforms, driving automation and orchestration through their networks and systems, all with a view to improving efficiencies and driving greater monetization opportunities. Despite significant efforts and investments, the needle has not moved sufficiently for most operators. There is a way forward, but it requires a shift to intent-driven processes that drive business outcomes and building a new infrastructure layer that is GenAI ready but more importantly, has a composable “agentic architecture”.

The end goal for telcos is an organization that is not just digital but cognitive, in the sense that the telco has a deep understanding of its customers but is also able to respond to them quickly and proactively. Predictably, there is a catch, which is that there is a critical need for converting the enormous amounts of network and customer data into something that is end-to-end, business-ready and easily consumable for analytics engines.

The need of the hour is a truly telco-vertical-specific data layer that cuts across domains and breaks or at the very least, bridges all of the data silos inherent within most telco networks today. Indeed, this legacy overhead and technical debt continues to hinder telco transformation and agility. Amdocs has a vision through its AmAIz suite and related offerings to bridge diverse data sources with any cloud and or data lake to enhance apps and services. Regarding data lakes, something many operators have invested heavily in, an interesting term we heard recently is that they have become “data swamps”. In other words, data is being collected but too little is being done with it, either due to it not being business-ready or simply a lack of AI smarts and expertise within the telco organization.

Telco relevance in the future will be driven by Agentic AI

Globally, telcos are all investing and, in many ways, are heavily invested in AI with a view to transforming their networks and making them more efficient. This is evident from AI trials, POCs and deployments across all domains from the RAN to the core of the network. AI is evident in the customer facing domains as well, with customer experience becoming a core mantra for telcos. Despite significant improvements in the customer-facing interfaces, it continues to be a slog behind the curtain as the diversity of systems that continue to be maintained have generated levels of complexity and technical debt that hinder agility.

Most operators represented on the TMForum autonomy frameworks barely at Level 2, still nowhere close to the Level 4 “highly autonomous” operations that is the aspirational benchmark. This is the chasm that needs to be crossed, or as Amdocs coined it, the “reality gap”. This is where agentic AI comes in as it plays a critical role in closing this gap. The Amdocs narrative is that building agents is one thing, but for true success, they must not be deployed as bolt ons.

To do this, two key success factors must be adhered to. One, agents that are deployed must be truly telco specific. In other words, agents that understand telco terminology and taxonomy of processes, cutting across domains as well as from network to service layers. Second, agents must be embedded into telco operations. In other words, agents that offload intensive work from humans as opposed to replacing them by using open APIs and protocols like Model Context Protocol (MCP) that will connect LLMs and SLMs to external tools, data and apps in a standardized way.

The roadmap to Agentic AI within the telco requires simplifying the telco OSS to become truly cross-domain, building a common data foundation that serves up business-ready data which is then consumed by agents for efficient network deployment and operations. The way forward for telcos of all stripes towards agentic AI is visible but with several twists and turns along the way.