The operator treats AI as a native network capability rather than an app layer add-on
In sum – what we know:
- Native network integration – Telefónica embeds AI features like transcription and virtual agents directly into its voice infrastructure rather than as overlay applications.
- Real-time capabilities – The platform offers near real-time call transcription with speaker identification and AI-generated summaries in over 100 languages.
- Strategic revenue shift – This move creates a new revenue line for the operator while deepening integration with customer business processes to reduce churn.
Telefónica is embedding generative AI directly into its business voice services in Spain — all of them. The rollout covers everything from basic corporate mobile lines to Centrex IP virtual PBX offerings and Centrex 365, the company’s cloud telephony suite tied to Microsoft 365. The idea is that any business phone call on Telefónica’s network can now be transcribed, summarized, and in some cases handled entirely by AI, without the customer bolting on any additional software.
That last part is what makes this different from the wave of AI contact-center tools already on the market. Microsoft, AWS, Genesys, and Salesforce all sell AI-powered voice features, but those sit at the application layer — over-the-top software layered on someone else’s phone lines, requiring integration work per customer and per channel. Telefónica is instead treating AI as a native network capability, baked into the voice infrastructure itself. If the customer opts in, every call becomes a potential AI-processed call by default.
The launch is initially aimed at Telefónica’s domestic enterprise market in Spain, though it fits into a broader group-level AI strategy that spans multiple operating companies.
Smart voice features and capabilities
The headline features are what Telefónica calls near real-time call transcription with speaker identification, plus AI-generated summaries that pull out key decisions, issues, and action items in formats customers can adapt to their own systems — CRM note templates, for example. Telefónica says the platform handles interactions in more than 100 languages, which matters for multinationals and public bodies serving diverse populations, though it’s worth noting that transcription quality across minority languages and regional accents tends to vary far more than the headline number suggests.
Then there’s “Centrex AI,” Telefónica’s virtual agent technology built on large language models. These are 100% virtual agents that can hold natural language conversations and execute tasks in back-end systems without a human involved — checking records, booking appointments, updating cases. The company is also promising omnichannel continuity, so a customer can start an interaction over WhatsApp and move to a voice call while the AI agent keeps the same context. That kind of channel-hopping is something CCaaS vendors have chased for years with mixed results.
The architectural distinction shows up here too. Rather than routing calls out to a cloud platform the way a Genesys or Amazon Connect deployment would, Telefónica’s AI agents join or take over calls through the operator’s existing signaling and media paths. In practice, that means enterprises keep their familiar phone numbers and PBX setups while the intelligence arrives via the network itself.
Network architecture and 6G readiness
Under the hood, Telefónica has built a modular AI voice agent platform combining LLMs, retrieval-augmented generation, workflow orchestration, and long-term memory — the standard agentic AI stack, but wired into places most AI deployments never touch. The platform seemingly integrates directly with IMS for call control and SIP interaction, and hooks into CRM and BSS systems for customer, billing, and subscription data. Audio from live calls is processed locally on AI-ready edge nodes deployed across Spain, which is how Telefónica plans to keep latency low enough for real-time transcription and conversational agents. Those edge data centers come out of the company’s work with Nokia.
That ties into the broader GSMA Open Gateway and CAMARA API push, and it’s a meaningful evolution from earlier efforts. Telefônica Brasil, for instance, previously used Microsoft’s Azure OpenAI service to power call-center assistants and speech analytics at Vivo. That was a useful deployment, but it was an application-layer project scoped to one contact center. The new approach generalizes those capabilities into a network-wide standard.
Spanish press coverage notes that Telefónica is currently the only operator in Spain offering unified call transcription and intelligent analysis across both fixed and mobile infrastructure. That’s Telefónica’s framing, but no domestic rival has publicly matched it, so the short-term edge looks real.
Use cases
The core value proposition is turning ephemeral conversations into structured data. Today, whatever gets said on a business call disappears when the call ends unless someone writes it down. With network-level transcription and summarization, that content can flow directly into CRM, ticketing, and analytics tools — commitments get tracked, sentiment gets measured, and contact-center agents stop spending chunks of their day on note-taking. For a large contact center, automated post-call documentation alone is a tangible productivity gain.
Telefónica is targeting large enterprises, medium-sized businesses, and public administrations, and the highlighted use cases are fairly concrete. Medical and retail appointment booking handled end-to-end by virtual agents. Citizen information lines for municipalities, where an AI agent fields frequently asked questions around the clock. Internal corporate lines where decisions get automatically logged rather than lost. Some of these will land better than others — a virtual agent booking a dental appointment is a well-bounded problem, while a citizen calling a public line with a complex or sensitive issue may not appreciate being handled by software, and automatic transcription of those calls raises GDPR questions Telefónica will need to answer clearly.
For Telefónica itself, the strategic logic is straightforward. Operators have spent years watching cloud vendors capture the value that rides on top of their networks while they sell minutes and bandwidth at commodity prices. Charging for AI processing embedded in the network is a genuinely new revenue line, and one that deepens integration with customer business processes — which makes switching carriers harder, for better or worse. It also gives the company a revenue-facing justification for its edge computing buildout. Whether enterprises embrace carrier-level AI or stick with the cloud players they already know, Telefónica has at least put a real product behind the industry’s long-running talk of monetizing the network.
Table of Contents
