Telkomsel Builds GenAI-Powered Incident Analysis System with Amazon Nova Pro, Cuts Response Time from Hours to Minutes

In today's fast-paced digital landscape, system downtime creates costly impacts that affect revenue, strain IT teams, and erode customer trust. For telecommunications providers, the financial stakes are particularly high—an industry survey found that high-impact outages in ASEAN countries have a median cost of about US$2.5 million per hour, 32 percent higher than the global average. For Telkomsel, maintaining service reliability above 99.9 percent is therefore critical to supporting over 150 million customers across Indonesia's diverse telecommunications market.

Recognizing that superior service reliability directly translates to customer satisfaction and business growth, Telkomsel's Cloud Center of Excellence team identified an opportunity to enhance its incident processing and management capabilities through AI and automation.

The telecommunications service provider set out to transform its incident management approach through GenAI and serverless computing. This initiative would reduce mean time-to-resolution, optimize resource allocation, and establish a foundation for scaling intelligent operations across the company's expanding cloud infrastructure.

To enhance operational efficiency and reduce incident response times, Telkomsel developed CELYNA, a GenAI-powered incident case analysis system built on AWS. CELYNA delivers incident analysis 21 percent faster than traditional methods, and its self-healing capabilities  accelerate resolution times by 83 percent. This improvement minimizes service disruption for Telkomsel's customers while optimizing resource allocation across operations teams.

Telkomsel developed CELYNA with Amazon Bedrock, a fully managed service to build and scale Generative AI applications, using Amazon Nova Pro, a multimodal foundation model that processes text, images, and documents to analyze complex incident data. With Amazon Nova Pro, Telkomsel’s engineers no longer have to spend up to 1 hour for each team analyzing logs across databases, applications, Kubernetes environments, and network infrastructure to identify root causes. Furthermore, instead of involving up to 20 people across multiple teams in resolution meetings for complex incidents, the engineers now use CELYNA for accurate root cause identification and actionable remediation recommendations.

With AWS Lambda, a serverless compute service, as the foundation for its serverless architecture, CELYNA can automatically scale compute resources based on incident volume while maintaining cost efficiency. After consolidating logs from databases, applications, Kubernetes environments, network infrastructure, and backbone systems into a unified pool, AWS Lambda processes this centralized data to provide detailed incident analysis across Telkomsel's hybrid cloud architecture. AWS Lambda's near real-time processing of the consolidated data triggers automated responses within minutes of detecting service degradation below the critical 99 percent threshold, allowing Telkomsel's engineering teams to focus on strategic initiatives rather than operational maintenance.

Building on CELYNA's success, Telkomsel plans to expand the system's capabilities to on-premises infrastructure, thus extending intelligent incident management across the company's entire technology portfolio. The team is also exploring additional AI services with AWS to enhance automation capabilities and support future digital transformation initiatives.

"CELYNA has fundamentally changed our operational approach, allowing our teams to maintain superior service reliability while optimizing resources. Our customers benefit from more consistent service availability, and our engineering teams can focus on innovation rather than reactive troubleshooting. With AWS, this intelligent automation represents the future of how we deliver excellence to Indonesia's telecommunications market," said Doddy Kristianto, Expert Engineer – IT Cloud, IT Corporate Solutions and Cloud CoE, Telkomsel.