Telefónica is integrating generative AI into its existing business telephone services, betting that the ordinary fixed and mobile call can become another source of structured enterprise data without requiring customers to move communications onto a separate AI platform.

The Spanish telecommunications group has added transcription, summarisation, and virtual-agent functions across services ranging from standard mobile lines to cloud-based switchboards. Large companies, public administrations, and medium-sized businesses are the intended customers, with Telefónica arguing that putting the technology into the network removes part of the integration work involved in adopting standalone AI communications software.

The offering can transcribe and summarise calls while turning conversations into information that can be used elsewhere in a business. Telefónica is also introducing Centrex AI, which uses language models to automate recurring interactions across voice and other channels and can connect with customers’ business applications.

Virtual agents support more than 100 languages and can operate continuously, according to Telefónica. The company says AI-assisted agents can manage tasks with 60% greater efficiency, while deployments have increased the volume of interactions handled by more than 10% and reduced average call-response times by 60%. Those figures come from Telefónica rather than an independent evaluation, so broader deployments will provide a better measure of performance across different workloads.

The approach moves enterprise AI into infrastructure that predates the current wave of generative software. Businesses already depend on telephone numbers, switchboards, call routing, contact-centre processes, and mobile connections, whereas many AI projects begin by asking employees or customers to adopt another interface. Telefónica is instead trying to make the communications layer itself programmable.

Telephone calls also contain information that often disappears once the conversation ends. A sales enquiry, support request, appointment booking, or citizen-services call can generate decisions that subsequently have to be typed into another system, while transcription and structured summaries can potentially remove some of that administrative work.

Telefónica has identified healthcare, public administration, retail, and industry among the sectors where it has worked on use cases. Examples include appointment management, citizen enquiries, bookings, and vehicle workshop scheduling, while local-government voice agents can answer common questions in several languages and route callers to the appropriate department.

The public-sector examples expose the boundary between processing a call faster and improving the underlying service. An AI agent can answer continuously and route requests more efficiently, but usefulness still depends on access to accurate information, reliable integration with departmental systems, and a route to a human when a request falls outside the automated workflow.

Turning conversations into structured data creates another governance problem for organisations that previously treated voice principally as a transient interaction. Transcription means information that may once have existed only during a call can become searchable and reusable, increasing its operational value while making retention, access, accuracy, and security controls more consequential.

Those questions become sharper when the technology is integrated into the communications network rather than bought as a separate tool for a narrowly defined team. Telefónica is offering the capability across traditional and cloud-based voice products, including fixed-mobile services and Microsoft-linked collaboration environments, so deployments could extend beyond contact centres into ordinary corporate calling.

That breadth gives the telecom operator a different position in the enterprise AI market. Cloud providers and software companies have largely approached generative AI through applications, developer platforms, or workplace assistants, whereas network operators control another layer through which organisations already communicate with employees, customers, and citizens.

Integration will determine how much value escapes the voice channel. A summary that remains inside a communications console saves less work than one that updates the appropriate customer record, triggers a follow-up task, or feeds an existing service process. Telefónica says Centrex AI can connect with business applications, placing those workflow links near the centre of the proposition rather than treating transcription as the finished product.

The operator’s performance claims now give customers numbers against which that proposition can eventually be judged. If organisations can reduce response times and process more interactions without simply transferring work elsewhere, the telephone network could become one of the quieter routes by which generative AI enters established business operations.