Login
Register
Luthfi Auzan, Vice President of Operations at IOH, details Indosat Ooredoo Hutchison (IOH) work on autonomous fault resolution.
IOH automates wireless access fault resolution with agentic AI
Who: Indosat Ooredoo Hutchison (IOH)
What: Shifting from human-driven, ticket-based operations and maintenance to AI-native, agentic operations
How: Deployed AI copilots and autonomous agents to automate fault monitoring, demarcation, remediation and field execution across mobile broadband domains
Results:
IOH is transforming the way it runs network operations by shifting from human-driven, ticket-based O&M to AI-native, agentic operations. Using a unified platform and a digital twin of the access network, IOH has deployed a new generation of AI copilots and autonomous agents to handle most fault monitoring, demarcation, remediation and field execution across mobile broadband domains.
The initiative centers on five core capabilities:
Together, these AI tools are helping IOH resolve up to 85% of common wireless access faults in a single step, according to the operator, while improving field productivity by 20% to 30%.
IOH operates one of the largest and most geographically complex networks in Southeast Asia. The operator manages over 55,000 base stations and more than 170,000 network devices supplied by over 15 vendors. The network generates around 5 million alarms every day and undergoes 200 to 300 configuration changes nightly.
External factors further complicate IOH’s operations. Indonesia’s extensive construction activity often causes more than 100 fiber cuts per day, while frequent power instability outside major urban areas causes repeated outages.
Historically, identifying whether an outage was caused by power, fiber or hardware required manual analysis and expert judgment.
“If humans do this manually, first it takes very long,” says Luthfi Auzan, Vice President of Operations at IOH. “Second, if you demarcate wrongly, instead of sending a team to the fiber cut location, you might send them to a power outage. That’s not really helpful.”
Before the O&M transformation, IOH’s operations teams dealt with alarm floods, siloed vendor tools and manual workflows. Field engineers regularly called backoffice teams for guidance, while multiple teams were dispatched in parallel to avoid misdiagnosing problems.
As network expansion continued at a pace of up to 4,000 new sites per year, IOH recognized that the model was unsustainable and that network growth could not continue to scale linearly with resources. The objective was not to flatten growth entirely but to make “the slope a little bit less steep”, according to Auzan.
IOH began its AI-native journey by focusing on fundamentals. In 2023, the operator unified data from multiple suppliers’ systems into a single operational platform designed to serve both human users and new “digital employees”, says Auzan.
On top of the unified platform, IOH developed a near-real-time digital twin of its mobile broadband access network, initially covering radio access network (RAN), microwave and IP – the domains responsible for roughly 80% of service-affecting incidents.
“The unified platform is the foundation for all the data needed to feed copilots and agents,” Auzan says. It stores topology, alarms, performance, history and workflow patterns, which the AI tools consume to make decisions and take action.
Organizational change was as important as adopting the new AI technology, according to Auzan. Rather than removing people from operations, IOH is transforming humans’ roles. Engineers whose manual tasks are being automated are reskilled to become AI builders, data engineers, policy engineers and orchestration engineers responsible for data governance, workflow logic and safe automation.
“[Engineers] are governing all the data, because garbage in will be garbage out,” says Auzan. “We need to ensure that someone in the back-end is governing the data, because wrangling and cleansing the data is 70% to 80% of the time and effort needed to make a meaningful use case for AI.”
IOH's fault monitoring and demarcation agent ingests millions of daily alarms and significantly reduces noise through intelligent correlation and pattern recognition. As a result, it says, raw alarm volumes have been reduced by up to 95%, enabling operations teams to focus only on actionable incidents.
The agent continuously monitors network conditions across RAN, microwave and IP domains, orchestrating specialized sub-agents for fault monitoring, demarcation and remediation. Leveraging a digital twin and graph-based reasoning, it analyzes complex cross-domain and multi-vendor network topologies, including ring and chain scenarios, to quickly identify breakpoints, dual-failure events and root cause network elements (NEs) through fault-propagation analysis.
This enables the AI agent to accurately distinguish between issues such as power outages, fiber cuts and hardware failures. By pinpointing the true source of faults, it eliminates unnecessary field dispatches and prevents duplicated troubleshooting across multiple teams.
Once a fault is demarcated, the domain-specific fault handling agent performs deep-dive diagnostics of the root cause. This agent executes targeted commands, validates outputs and uses retrieval-augmented generation (RAG) to recommend resolution actions and standardized fix procedures. It communicates with the fault monitoring and demarcation agents via Model Context Protocol (MCP), and recovery actions are executed automatically where feasible.
In parallel, the system orchestrates the full lifecycle of trouble tickets, from automated dispatch to the appropriate team, through progress tracking and restoration validation, to final closure once service is confirmed.
Human intervention is required only for exceptions or when physical on-site work is unavoidable. This automation has significantly reduced operational handoffs, shortened mean time to repair (MTTR) and ensures consistent, policy-driven execution across operations.
As a result, Auzan notes, up to 85% of common wireless access faults can now be resolved in a single step without human involvement, marking a clear shift from traditional ticket-based operations toward a human-machine collaborative model.
IOH has put in place a front office (FO) copilot assists the front-office team in handling exceptions reported by the Fault Handling Agent. It provides interactive job assistance to help the FO quickly handle the exceptions.
In addition, the FO copilot provides the status summary before and after the network changes, greatly reducing the time required for the front-office team to check the network change status from about 10-15 minutes to just 2-5 minutes.
Instead of navigating multiple tools, engineers receive prioritized insights and decision support, improving both speed and accuracy. By maximizing the efficiency of human-machine collaboration, network operations efficiency in the DIOC has improved by more than 20%, while availability and traffic levels have reached record highs, according to Auzan.
The FME Copilot represents one of the most visible changes to IOH’s day-to-day operations. Previously, engineers in the field routinely called back-office teams for guidance and confirmation at each step of an on-site visit. The copilot provides step-by-step troubleshooting guidance, allowing field maintenance engineers to simply enter natural language queries.
In the background, the system executes real-time CLI commands on the relevant network nodes and returns summarized results to the user. It can check configurations, verify parameters and status, confirm alarm clearance and validate service restoration.
This significantly simplifies troubleshooting workflows, enabling faster and more consistent issue resolution while reducing reliance on manual command execution and deep domain expertise.
The impact on productivity has been substantial, according to Auzan, with the copilot replacing nearly 90% of interaction between the field engineers and humans in the back office. representing a 20% to 30% onsite efficiency improvement. More accurate demarcation also ensures that engineers are sent to the right location with the right instructions the first time.
By adopting agentic operations, IOH has begun to decouple network growth from operational headcount. Around 20% of operational work is already handled by digital employees, with AI agents taking over repetitive tasks. Over the next two to three years, the company expects this proportion to grow significantly as zero touch operations expand.
Rather than reducing staff, IOH has focused on workforce transformation. Hundreds of employees have been reskilled into AI operations specialists, data engineers and policy designers.
“We don’t want people doing repetitive Excel work anymore,” Auzan says. “We want them doing higher-value work. Agents and copilots handle the repetition.”
IOH’s long-term vision is an “agent everywhere” operating model. The operator plans to deploy agents across other processes including planning, deployment, optimization and service assurance.
“In the next two or three years, operations will drive the technology,” Auzan says. “Operations handles the data, and planning becomes the user of that data.”
As AI agents mature, IOH expects operations to move from a reactive function to a strategic driver of network evolution, using real-time insights from the digital twin to shape planning, investment and customer experience.
“The change is inevitable,” Auzan concludes. “This is the mindset we want to encourage."
