SK Broadband overhauls network quality management with AI agents

SK Broadband is transforming how it manages network quality by building an environment where any employee can create and deploy AI agents in their day-to-day work.

The company said Monday it launched an in-house development platform called Playground in February, equipping it with network data analysis and coding support tools. Network teams and the company's AT/DT Centers jointly developed the platform.

SK Broadband said employees had long struggled to bring data analysis or automation into their work because of technical barriers. Using AI on the job required navigating complex procedures — applying for server access, setting up development environments and installing software libraries — before any practical work could begin.

Playground was built to remove those obstacles, the company said. It automates environment setup that previously took more than two months, cutting the time to start development to five minutes.

Playground can also be used alongside LDAS, the company's location-based internal data analysis system. Employees can draw on a range of internal data — including network equipment metrics, quality indicators, traffic data and customer experience index (CEI) figures — when building AI agents.

About 600 AI applications have been developed and are running through Playground, SK Broadband said. Of those, roughly 30 are fully autonomous AI agents capable of assessing situations and taking action on their own, and all are already deployed in the field.

The AI control and diagnostics agent C-One is the flagship example. Drawing on CEI data, C-One automatically detects anomalies in the wired network, identifies root causes and ranks inspection priorities. It can also pinpoint sites that need attention, generate reports automatically and send them to the responsible staff.

In practice, a network manager responsible for the Gangnam area of Seoul can use the C-One dashboard to check at a glance the average internet quality score for their zone and the number of recent outage reports. Priority sites — broken down by neighborhood and building — are displayed on a map. Selecting a building and running an AI analysis prompts the system to synthesize anomaly signals, including quality score trends, complaint volumes and optical signal strength, then recommend follow-up actions such as remote equipment resets or on-site fiber line inspections.

SK Broadband plans to upgrade C-One into a fully autonomous recovery agent capable of handling everything from fault detection through resolution and restoration without human intervention.

The company also operates a system in which servers deployed nationwide send signals directly to subscriber devices to measure Wi-Fi and high-speed terminal quality in real time. Devices with poor quality are automatically flagged, and AI then analyzes fault histories and time-of-day patterns to diagnose problems, estimate causes and suggest actionable fixes — allowing the company to address potential failures before they affect service.

"The way our people work is fundamentally changing as they build AI agents themselves and put them to use in the field," said Seong Jin-su, head of SK Broadband's Network Center. "We will continue to spread a culture of AI development and keep raising network quality."