While many sectors use AI mainly for back-office tasks, telecommunications involve live infrastructure where problems can escalate quickly and manual coordination often fails.
As 5G densification, growing traffic, and increasing service complexity challenge old operational models, operators are bringing AI agents into production. This move helps stabilize networks, cut operating costs, and protect profit margins.
AI agents in telecom are used carry out workflows that cross various operational areas. They monitor real-time network data, spot issues, link data from radio access networks and core infrastructure, and initiate fixes without waiting for human approval. These agents work within live systems, rather than on the organization’s periphery.
According to a report from Google Cloud, 56% of telcom executives reported their organizations are actively using AI agents in production, with nearly half (43%) saying they have already launched 10 or more. Notably, 1 in 5 respondents also said these agents are deeply embedded across their operations.
Agents in Core and Customer Workflows
For example, Deutsche Telekom deployed an AI-powered RAN Guardian Agent that continuously monitors radio network performance, detects anomalies and autonomously initiates corrective actions. The operator said this system reduces the time required for diagnostics and corrective tasks from roughly an hour to just minutes, improving response times and reducing reliance on human intervention.
Telefónica has implemented AI agents for closed-loop network control, aiming to maintain stability during traffic spikes. The agents process data from core network elements, forecast capacity constraints before they can degrade service, and automatically adjust routing policies or allocate more computing resources. Tasks that used to need manual actions from network operations staff now run automatically, allowing engineers to concentrate on capacity planning and system upgrades.
AT&T also announced the expanded its use of agentic AI to include customer-facing systems. The company is using agents to manage account updates, billing questions and service requests by accessing customer relationship management systems, billing platforms and service activation processes. AT&T also leverages AI agents in network engineering to examine traffic patterns, suggest network changes, and simulate the effects of configuration modifications before they are applied. They reported quicker resolution times for customer service requests and improved accuracy in network planning.
Adoption Obstacles
However, telecom operators still face challenges that slow down the adoption of AI agents. According to IBM, integrating and managing complex data is the top barrier to AI adoption for telecom operators, cited by 67% of respondents, followed by legacy IT infrastructure, cited by 52%. As a result, many operators lack the data platforms and engineering resources needed to train AI systems internally, pushing them to partner with cloud providers that offer pretrained models tailored for telecom use cases. Rollout timelines can extend further as operators validate agent performance across multiple network domains before granting full production access.
Most carriers operate on legacy systems built over many years, with infrastructure that wasn’t designed to support real-time APIs or automation. As Salesforce mentioned integrating AI agents into these systems is complex requires additional middleware, infrastructure upgrades and governance frameworks to define what actions agents can take without human oversight.
Despite the challenges, telecom executives remain optimistic about generative AI’s impact. PYMNTS Intelligence previously found that 67% of telecom executives believe generative AI can improve IT service provision, while 85% see strong potential for AI to positively affect both operations and network performance.
Telecom organizations also report measurable gains from deployment, according to Google Cloud study. The company found that 72% of respondents reported increased productivity in IT workflows and 55% cited gains in non-IT workflows. Executives also reported faster time to insight, at 58%, and improved accuracy, at 55%. Security benefits were also cited, with 82% reporting improved threat identification, 72% stronger threat intelligence and response, and 58% faster time to resolution.
Source: https://www.pymnts.com/