AIOps builds scalable, AI-driven network operations today

[Monica] Why do we need AI and AIOps

Monica: Welcome to this webinar on how AIOps can build scalable, AI-driven network operations today, and delve into how AIOps has already been deployed commercially at Claro Colombia, how the deployment will continue to expand and evolve over time, and what this deployment teaches us about how AI and AIOps change the way we operate telecom networks.

Operators are very good at running operations and running networks, but there are still large pockets of operational inefficiency and money left on the table. There is a wide scope for improving network efficiency and customer experience, and AI is the most effective tool to achieve this.

In this context, adopting AI is no longer optional. It is not a nice-to-have or something for the future. It is now necessary to improve efficiency. While the expectations on what AI can deliver are high, there is also a pervasive that progress is slow and integration of AI in the networks challenging, if not downright dangerous.

The graph shows the overall enterprise market and not specific to telecom. It shows how almost everyone is adopting AI, but only 7% of enterprises have full-scale deployments. Most of them (62%) are experimenting or piloting AI. This is not odd or problematic because AI is still a new technology in its early adoption phase. In telecoms, we face a similar situation as we make the first steps into AI territory.

What makes today’s webinar different is that we are talking about one of the commercial deployments, firmly in the top 7% dark purple area of the chart.

We discuss why operators want to deploy AIOps in the first place, what opportunities it creates, what challenges exist, and what the initial results look like. I emphasize “initial results” because we are still in the early days. Similarly to 5G and its evolution to 6G, there is an incremental adoption path, with each operator moving at its own speed and following a different strategy.

AIOps is not an add-on module superimposed on the current operation platforms. It is a new execution and control layer for automation, built into the TM Forum automation framework. We should think of AIOps as the combination of the computational power of AI and unparalleled depth of human expertise. Only when these two elements work together will we reap the full benefits of AI.

I am privileged to have excellent speakers with us today. We have Carolina Garcés Zapata, Service Operations Center Manager at Claro Colombia. She is responsible for service assurance and operational administration, ensuring that the network operates correctly and that customers are satisfied. We also have German Perez-Trozzi, Vice President of Product Marketing at Iquall Networks, who focuses on network automation, and Fabian Aranda, AIOps Solution Architect at VIAVI Solutions. They have been working together for some time on this initiative, which was featured as a catalyst project at the 2025 TM Forum conference in Copenhagen. The next phase of the project will be showcased at DTW 2026 in June.

We will start with Carolina, who will explain what Claro has been doing. Then Fabian and German will discuss AIOps and VIAVI and Iquall’s solutions. We will conclude with Q&As and concluding remarks.

[Carolina] AIOps at Claro

Carolina: Communications service providers are operating in an environment of high complexity. Networks are evolving rapidly. Traffic demands continue to surge, and customers expect seamless, real-time service experiences across mobile, fixed, and converged networks.

At the same time, operators must manage increasingly diverse architectures, cloud-native deployments, and tighter regulatory expectations. Internally, organizations face their own pressures, including cost constraints, demands for higher operational efficiency, and skills shortages that make traditional manual operations no longer sustainable.

While operators have started adopting analytics, automation, and AI driven assurance, these initiatives often remain fragmented. They improve isolated domains but fall short of delivering true end-to-end operational transformation.

In this webinar, we will explore these operational challenges, examine the solutions emerging today, and discuss the steps required to move toward more autonomous, scalable, and resilient network operations. I will talk about how end-to-end incident automation can create a bridge between network performance, customer experience, and operational efficiency.

Before looking at the solutions, it is important to understand the problem we are trying to solve.

Today, CSPs face what can be described as a perfect storm: rapidly growing complexity, an explosion of devices, diverse SLA-driven services, and limited end-to-end visibility across the network.

Traditional KPIs often fail to reflect the real customer experience. Data remains siloed across different platforms and domains, making cross-domain correlation difficult, slow, and highly manual. This leads to delayed issue detection and slower remediation.

As a result, mean time to repair (MTTR), customer-perceived service degradation and churn increase, while net promoter scores (NPS) decline. Regulatory bodies are adopting new crowdsourced measurement frameworks, and operators have to comply with experience-based standards and provide metrics for both network KPIs and experience-based metrics. Operations must now comply with experience-based standards, not just network-based metrics.

Operational efficiency is compromised as well. Manual processes increase errors, delays, and operating costs. Expert knowledge becomes a bottleneck for multi-domain correlation, and incomplete insights can drive poor investment decisions, increasing capex and opex. These challenges highlight the need for a fundamental shift from reactive to proactive, automated, and customer-centric operations.

Our solution unifies customer-centric insights, AI-driven automation, and cross-domain visibility. With advanced AI and machine learning, we detect network degradation before it affects our customers. We automated root-cause identification and resolution, and require minimal human intervention. A key differentiator is our integration of real customer experience data with traditional network telemetry. Instead of viewing network performance in isolation, we correlate it with what customers actually experience, prioritizing incidents based on real impact rather than purely technical alarms. We also generate a multi-domain, multi-service, multi-vendor topology to provide full end-to-end service tracing, identifying failure points that may affect availability and performance.

On top of this, our closed-loop incident management strategy enables fast, effective, and cost-efficient remediation, forming a scalable foundation for next-generation autonomous operations.

The business impact of this end-to-end automation strategy includes measurable benefits such as faster time to resolution, reduced customer impact, and consistent operations at scale. By combining proactive detection, AI-assisted remediation, and customer-based prioritization, we decrease false alarms and MTTR, and target network investments more effectively.

We have seen up to a 67% reduction in operational costs for the network operations center (NOC) and the service operations center (SOC). Automation minimizes manual effort, reduces dependence on external providers, and optimizes resource allocation.

With end-to-end visibility and closed-loop automation, operations become more resilient and scalable, supporting increasingly digital and experience-driven services.

The graph lists the strategic next steps that will allow us to expand automation and strengthen our ability to detect, prevent, and resolve issues with greater precision and speed.

The first pillar is anomaly detection. By identifying unusual patterns or deviations from expected behavior in real time, we can detect early indicators of service degradation before they have a negative impact on customer experience. This gives us the ability to react faster and with higher accuracy.

The second pillar is predictive detection, where AI and machine learning models help us anticipate potential issues before they occur. Instead of diagnosing problems after they happen, we can intervene proactively, protect service quality, and minimize disruptions, especially during high-demand scenarios.

The third element is about taking action at scale. Virtual agents automatically execute predefined resolution steps, validate conditions, and update workflows. This significantly reduces manual effort, minimizes errors, and accelerates incident handling, allowing our teams to focus on higher-value activities.

Finally, proactive customer communication enables us to inform users about potential service impacts before they experience them. This improves transparency, strengthens customer trust, and reduces support calls.

Together, these capabilities form the foundation of Claro Colombia’s next phase of operational evolution, where automation, intelligence, and customer-centricity come together to deliver a more resilient, efficient, and proactive network operations model. This path that will enable us to scale, adapt, and continue leading in service quality and operational excellence in Colombia.

[Fabian] How does AIOps enable network autonomy

Fabian: I would like to put Claro Colombia’s AIOps adoption into a wider context, by explaining how TM Forum standards provide a common roadmap for service providers that want to modernize their operations in a structured and realistic way, through three concepts: AIOps, closed-loop automation, anomaly detection and resolution (CLADRA), and autonomous networks.

AIOps uses AI and advanced analytics to process large volumes of operational data, including alarms, metrics, logs, topology data, and inventory. It automatically detects issues, correlates root causes, and can predict problems before they occur. I want to emphasize that AIOps is not only analytics. It is a decision engine that fits into closed-loop automation and drives actions in the network.

This is where CLADRA comes in, using closed-loop automation to connect data, analytics, decision logic, and policies to ensure that automation is not a black box, but rather something that is governed, explainable, and, most importantly, repeatable.

The third concept is the autonomous networks maturity model. When we talk about autonomous networks, we are not talking about turning everything into AI all at once. The TM Forum framework defines clear maturity levels, from level zero to level five, describing how automation evolves over time.

AIOps makes actionable the massive amount of data that telecom networks generate every day. Service providers deal with enormous volumes of data from inventory systems, topology systems, alarms, metrics, and logs across multiple domains and vendors. Humans and traditional tools cannot process this data in real time in an efficient way.

AIOps uses machine learning and advanced analytics to continuously analyze data and transform it into actionable insights that can trigger ticket creation, automatic actions, or workforce activities in the field. AIOps automates the end-to-end process by enabling automatic ticketing, monitoring, alarm correlation, anomaly detection, forecasting, and identification of the most likely root cause of an incident. Root-cause identification is one of the most important elements for reducing MTTR. In turn, this reduces operational noise, shortens troubleshooting time, and accelerates resolution.

AIOps also enables the shift from reactive to proactive operations that Carolina described earlier. Instead of waiting for alarms or customer complaints, AIOps can predict service degradation by comparing normal network behavior with abnormal behavior. This allows operators to resolve issues before service is affected, directly improving customer experience and SLA compliance.

AIOps reduces manual effort by automating repetitive tasks and guiding engineers toward the correct actions. When integrated with closed-loop automation, AIOps not only detects problems but also recommends actions that can be executed automatically.

CLADRA describes how automation works within autonomous networks. Rather than relying on isolated scripts or standalone automation tools, CLADRA provides a structured and standardized approach for building closed loops that operate across multiple domains, technologies, and vendors.

CLADRA enables automation to be predictable and repeatable. Data is continuously collected from the network, analyzed, and evaluated. Decisions are made based on intent and policy, and actions are executed automatically. The outcomes are fed back into the loop to validate results and enable continuous learning and improvement.

Because each step in the loop is well defined and governed, operators can trust automation behavior and scale it safely. Ultimately, CLADRA reduces the need for manual intervention in network operations, and the time required to detect and resolve issues across the network. It is one of the most powerful mechanisms for reducing operational expenditure because it automates the full detect-to-resolve lifecycle.

In addition, CLADRA enables a shift from reactive to proactive operations. Closed loops can detect degradation and trigger corrective actions automatically.

Closed-loop automation is a core capability required to reach autonomous network maturity level 3. At this level, service providers are no longer experimenting with automation. They are operating continuous, cross-domain closed loops that deliver measurable, positive business impact.

[German] The path to network autonomy

German: The aim of the autonomous networks’ maturity model is not to rank operators, but to provide a common and structured language to discuss network autonomy in real, complex environments. Autonomy develops progressively across three core dimensions: the level of AI and automation, how intent is defined and enforced, and the role of humans in the operational loop.

At autonomous network level 2, operations are assisted. Automation platforms continuously ingest metrics and events received via telemetry from the network. These inputs are correlated to provide operational context, and closed-loop automation already exists at this level. When predefined conditions are met, the network executes deterministic remediation actions. Behavior is predictable and rule-based. There is no intent evaluation and no autonomous decision-making. Humans design, validate, and maintain automation, but execution does not require human intervention.

At autonomous network level 3, autonomy becomes conditional. AI is introduced to classify situations as service-impacting events and to evaluate the operational context. Actions are validated against explicit service intent and operational guardrails, including SLA thresholds and policy constraints. The actions themselves do not change from level 2. What changes is how decisions are made. Humans intervene primarily for exceptions, escalations, or policy-based approvals, not for routine operations.

At autonomous network level 4, the network reaches high autonomy. Intent becomes the primary control mechanism. AI operates with full network topology and service context, combining real-time telemetry with inventory and configuration data. The set of possible actions remains similar, but the decision model evolves. Humans are no longer part of real-time execution. Their role shifts to governance, policy definition, and continuous improvement.

You may notice that level 1 and level 5 are not included here. Level 1 represents basic scripting and isolated automation. It is a necessary foundation, but it does not include intent or autonomous decision-making.

Level 5 represents a fully autonomous, self-evolving network at the ecosystem level. Today, this level remains largely conceptual and not achievable in a real brownfield environment. It is a North Star.

Autonomy is measured differently across network domains, including the radio access network, fixed access, mobile core, IP core, and telecom cloud. Because of time constraints, I will focus on a simplified example in the IP core to illustrate the main idea.

At level 2, the network detects degradation by correlating metrics and events received via telemetry from traffic data. It then executes predefined deterministic remediation actions, such as changing a border gateway protocol route weight, shutting down an interface, or applying a predefined QoS policy change.

At level 3, supervised AI evaluates service intent, SLA impact, and operational guardrails, and decides whether those same actions can be executed autonomously.

At level 4, the network, driven by AI–based decision-making, optimizes the IP core autonomously. It evaluates multiple options and selects the optimal action based on business priorities, SLA objectives, and cost constraints.

[Fabian] How VIAVI AIOps solution helps telco service providers

Fabian: One of the biggest challenges that telco service providers face when adopting AIOps is understanding which components enable AI-driven operations without adding unnecessary complexity or overlap.

With its modular architecture, AIOps can fit into different ecosystems, where some components may already exist, and so that it can be deployed in phases and support progressive maturity evolution.

The first AIOps component is network inventory. An AIOps network inventory establishes a reliable, up-to-date record of physical and virtual network resources, as well as their logical and physical relationships. It becomes a single source of truth, which is critical for accurate analytics, correlation, and automation.

On top of inventory, we have network topology, which provides end-to-end visibility. AIOps network topology maps how network elements connect across multi-domain and multi-vendor environments, for example, from the radio access network through transport and core networks. By combining inventory, configuration, and performance data, AIOps enables real-time visualization of the full-service path and identification of potential service degradation.

Next is network performance assurance, which adds near–real-time intelligence to analyze KPIs across different domains, detect degradation, support troubleshooting, and enable continuous network optimization.

This feeds into service quality assurance, where raw KPIs are translated into service-level KPIs. It allows operators to track service quality rather than only network quality and to understand how performance affects customers.

We then have network fault management, which reduces operational noise on events by eliminating duplicated alarms, correlating alarms, and accelerating root cause identification.

This information flows into the incident management component, where events are transformed into actionable incidents and enriched with full contextual information to support resolution.

Machine learning and network automation close the loop. Machine learning enables anomaly detection and forecasting, while automation workflows trigger notifications and corrective actions. Together, these components enable end-to-end service visibility, proactive management, and closed-loop automation.

Finally, network orchestration pulls everything together.

[Geman] Orchestration completes the AN journey

German: At this point, the question is how to make this journey work in a real network. A real telecom network is highly complex and must remain operational while changes are introduced. Downtime is not an option.

This is where AI-driven network automation and orchestration at scale, combined with governance, become critical. Iquall’s Multipurpose Automation Tool (MAT) framework acts as an orchestration layer between observability, business systems, and the network itself.

The graph shows observability, with telemetry APIs, alarms, logs, probes, and metrics coming from multiple network domains. Solutions such as VIAVI Nitro play a critical role in collecting, normalizing, and simplifying this data at scale, providing high-quality, near–real-time network insight.

The observability layer is essential, but it is only one aspect of end-to-end architecture. The next challenge is to turn insight into controlled, predictable, and governed action. This is where MAT comes into play. MAT abstracts network complexity across multiple domains and vendors, providing a consistent execution and governance layer through workflows, closed-loop policies, CI/CD pipelines, and controller configuration management. All of this takes into account the real-time network state and operational constraints.

At the north of the architecture, the platform integrates natively with business systems such as ITSM, CRM, billing, BI, WFM, and AI agents. It leverages industry standards such as TM4, Open ABI, TM Forum, MEF, and MCP while also providing flexibility to build custom APIs to support legacy systems and in-house development.

Together, these elements enable closed loops that start with high-quality observability and end with governance execution at network scale. Operational actions are aligned with service and resource intent and with business objectives.

MAT combines telecom-native automation capabilities with self-service features for building and evolving use cases. This allows operators to start quickly with proven use cases while retaining full control to adapt and extend automation as the level of autonomy grows.

Platforms such as MAT are not about automating individual tasks. They enable operators such as Claro Colombia to move from assisted operation to AI-driven governance, closed-loop production environments.

More importantly, this approach changes how operators work. Processes are redesigned from the ground up, with platforms like MAT allowing for a faster, safer and more scalable deployment of automation.

[All] Q&A and concluding remarks

Monica: Thank you Fabian, German and Carolina for your excellent presentations on how VIAVI and Iquall supported AIOps deployment at Claro Colombia.

We now have a few questions, and I would like to start with Carolina. What did you find surprising as you embarked on this project? What did you learn during the process of deploying AIOps in a network, moving from architectural design to real production in a complex, multi-domain environment?

Carolina: What surprised us the most was how different things look once automation reaches the real world. On paper, everything appears clean and logical, but in production, you deal with data gaps, legacy systems, inconsistent processes, and teams working at different speeds.

We quickly realized that the technical architecture represented only half of the challenge. The other half was aligning people, processes, and data across many domains. Once that alignment was achieved, automation performed exactly as expected. However, reaching that point requires strong collaboration and adaptability.

Monica: German, as Carolina said, when you move from theory or design to reality, many unexpected issues emerge. Does something similar happen when moving from standards to actual solution development? Standardization and openness sound appealing, but they can also be intimidating. Do they really matter when we talk about artificial intelligence and scaling?

German: Standards and openness are necessary because they make automation and AI scalable and sustainable, not just deployable in a single project. They allow operators to decouple network automation from vendors, domains, and specific tools, and to avoid lock-in. They provide the governance, consistency, and repeatability that are mandatory at telecom scale.

Open standards allow automation and AI to evolve over time without re-architecting the network every time a new use case is introduced. Without them, automation remains fragmented and difficult to scale, and thus unsuitable for telecom environments.

Monica: Let me take this one step further. Standards create a foundation that can be applied across multiple operators. Fabian, what lessons did you learn from the Claro deployment that can be generalized to other operators that may differ in size or operational needs?

Fabian: One key lesson is the importance of establishing the basics first. This means consolidating data from multiple sources across different network domains and vendors. This is the foundational step in the journey. It is also critical to simplify existing operational processes. If inefficient processes are automated without simplification, automation will only multiply those inefficiencies.

Once the basics are in place, the next step is to optimize the processes. After that, the transition from maturity level 2 to level 3 requires full closed-loop automation. This is where real efficiencies emerge, manual effort is reduced, and operations become simpler and deliver better results.

Monica: We have discussed technology extensively, but there are also people involved in this process. Teams must work together and operate differently, as Carolina pointed out earlier. Carolina, how has AIOps changed the way your internal teams collaborate?

Carolina: Once AIOps began providing insights based on real customer impact rather than only network metrics, collaboration changed significantly. Network teams, information technology teams, and customer operations teams started sharing the same information and speaking the same language. Instead of focusing only on their individual domains, everyone began prioritizing the overall customer experience.

This led to faster decision-making, smoother collaboration, and a more unified approach to problem resolution.

Monica: Your experience at Claro suggests that increased automation brings human collaboration and expertise to the forefront. There is often a narrative that automation makes networks less personal and humans less relevant, but you suggest it is changing the nature of and approach to their work.

German, you described different levels of autonomy earlier. How do operators move from one level to the next?

German: Our experience with Claro Colombia shows that autonomous network maturity is as much a cultural transformation as a technical one. Our role at Iquall is not to impose a predefined recipe, but to work closely with our partners such as Claro Colombia to align automation with their specific business needs, network realities, and operational constraints in a true brownfield environment.

New technology trends emerge every year and are often presented as universal solutions. Our focus is not on trends, but on what delivers measurable improvements in network operations for our partners.

This means building solid, iterative solutions, learning from errors, and moving quickly toward production deployment.

Too often, service providers spend more time debating which use cases to pursue than executing them. That has not been the case with Claro Colombia and VIAVI. The focus has been on fast execution, improved customer impact, and higher efficiency.

Monica: You mentioned customer focus, which can be lost during complex technical implementations. Fabian, how do operators move from raw data and metrics to an understanding of customer perception using AIOps?

Fabian: This is an area where Claro did an excellent job. The key was integrating customer-measured data, including crowdsourced data, into the AIOps solution. This data was analyzed alongside network inventory and topology to identify network issues that affected customers. Once issues were identified, actions were directed at the appropriate teams, avoiding delays and finger-pointing between groups.

By fully automating this process, operators can resolve issues faster and more consistently. In the next phase of this project with Claro Colombia and Iquall, we expect even stronger results.

Monica: What potential mistakes should operators avoid when adopting AIOps?

Fabian: Operators should prioritize the application of AI where it delivers the greatest impact, particularly in areas that reduce manual effort significantly. It is also essential to follow a standardized framework to ensure scalability and a common language across organizations.

Reducing complexity and keeping solutions as simple as possible is critical. Complexity often grows without delivering proportional value.

Finally, operators should choose not just a solution, but an automation partner. This partner should support the entire journey, not simply provide a tool and leave the operator to manage the rest.

German: I worked for fifteen years at a service provider, so I understand how challenging this is. Operators face declining average revenue per user, rising costs, ongoing training requirements, and legacy equipment that cannot simply be shut down. This requires building a strong ecosystem of partners and working together transparently and closely.

Errors must be treated as part of the learning process. This does not mean causing outages or impacting customers, but allowing teams to experiment, learn, and iterate. Scaling automation requires many people working together. This is not a small project limited to a single team. It involves technology, operations, and information technology teams across the organization, all learning and evolving together. Moving quickly and learning iteratively is often more valuable than attempting to do everything perfectly from the beginning.

Carolina: For me, the key to avoiding mistakes is aligning the business with the appropriate costs and investing in the tools, time and resources required to implement solutions that are sustainable for the business.

Monica: Claro Colombia’s adoption of AIOps stands out as one of the earliest commercial deployments where automated operations and advanced observability led to a powerful shift from reactive to proactive operations. Claro Colombia has been able to leverage massive amounts of data to make informed decisions and ultimately improve customer experience.

This work is grounded in the framework developed by the TM Forum for autonomous networks, which provides operators with a structured and practical path forward through the five-layer autonomous networks maturity model.

However, progressing to higher autonomy levels is not enough on its own. Network orchestration must also be introduced to ensure that all these elements work together effectively.

This integration is required not only to deliver better performance and improved customer experience, but also to create tangible business value. This includes reducing operational expenditure, lowering MTTR, and improving overall operational efficiency.