Enhanced network management: unleashing the power of AI


In network management, artificial intelligence (AI), used effectively, has the potential to make operations more efficient, reduce costs, mitigate downtime, and help focus investments where they are needed most. What do you need to know? Below we give you some progressive example use cases Orange Wholesale International has implemented in our networks and some feedback from our own experience.

Making network management simpler and more efficient is an ongoing challenge. As networks become smarter and more complex, keeping them running smoothly and problem-free comes with new barriers to overcome, and takes on greater urgency. Companies need to prioritize it: research shows that the average cost of system downtime for enterprise networks is $300,000 to $400,000 per hour

It’s something Orange Wholesale International has been investigating to help drive greater efficiencies in our networks and give our customers greater resilience, reliability, and trust in their networks. Below, we outline some example use cases where we’ve used Artificial Intelligence (AI) to drive significant network management benefits for customers.

Infrastructure optimization

Maximizing infrastructure usage is always a key goal when enhancing network management. It encompasses things like capacity planning and routing to drive cost efficiencies and reduce wastage. Capacity planning is an essential element in network management and infrastructure optimization. Proactively addressing potential bottlenecks before they occur means you can deliver optimal network performance and minimize downtime for customers. It enables better-informed decision-making around where and when to invest in network infrastructure and helps you forecast future network capacity needs by analyzing historic network traffic data. 

Real-time routing is another area where efficiencies can be had. We use data-driven approaches to power real-time decision-making on CDN overflow, or to automatically select the best route for voice traffic. This approach lets us provide the best quality of service (QoS) with the lowest cost for our clients. 

More efficient IT operations with AIOps

AI for IT Operations, or what Gartner calls AIOps, can help mitigate the challenges IT teams face in managing increasingly complex IT environments and keeping networks running at peak performance.

In network management terms, AIOps can be a powerful tool in troubleshooting and ticketing. For troubleshooting it begins from the point of network infrastructures today being increasingly complex. They’re typically built using a combination of hardware and software from multiple different providers. As such, alerts and logs are multiplied, and so in amongst all that noise, it becomes difficult to detect weak signals.

AI can analyze different logs and group logs according to established patterns, identify weak signals, and proactively initiate troubleshooting. Based on previous incidents, AI can identify possible causes and further accelerate resolution. Predictability is central to this. AIOps can work with legacy IT systems and business applications like ERP, and gather together data that was previously locked in siloes. This means it can provide a regularly updated, accurate, and synchronized view of IT operations, enabling IT staff to spot and react to pertinent issues in real-time.

AIOps is a burgeoning area in network management: research has found that 94% of companies believe it is important or very important for AIOps to manage network and cloud application performance moving forward. At Orange Wholesale International, we use AI and algorithms to help our technical experts refine alerting thresholds to reduce the “noise” and allow operational teams to focus on real alerts.

The predictability offered by AIOps offers benefits for ticketing, too. AI can analyze words used by customers when reporting incidents and enable tickets to be categorized more precisely and prioritized more effectively. This helps the Service Management Center (SMC) automatically redirect tickets to the most appropriate Technical Management Center, which has a direct impact on resolution time.

Fraud detection

Worldwide digital fraud losses have been forecast to exceed $343 billion between 2023 and 2027. AI can analyze vast amounts of records and traffic, and detect patterns and anomalies that may indicate fraudulent activities. AI-powered fraud management systems can identify and prevent various types of fraud, such as payment fraud, identity theft or phishing attacks. They’re also able to adapt and learn new fraud patterns and trends, thereby improving their detection abilities over time.

For several years, Orange Wholesale International’ Hubbing solution has leveraged the power of AI to proactively analyze, detect, and block fraudulent calls through continuous monitoring, preventing International Revenue Share Frauds (IRSF). Our anti-fraud solution is based on Khiops, the innovative machine learning (ML) solution developed by Orange. It speeds up information analysis, data preparation and definition of fraud detection models. Anti-fraud experts can then speedily identify potential fraud and proactively block it.

Identifying business opportunities

AI’s data analysis capabilities and proactive, predictive nature mean it can help uncover inefficiencies, patterns, and abnormalities that can deliver cost savings or revenue growth.

In the case of network management, AI could identify underutilized network resources that can be optimized or provide insights into customers’ usage patterns, enabling operators to personalize offerings and bring in new revenue streams. At Orange Wholesale International we’re using AI to proactively predict any potential overloads on customer links, so we can propose an upgrade before congestion occurs. We’re also using it to predict when a commit may not be reached and to detect imbalances in traffic with peers. Because a commit is usually negotiated in exchange for a lower tariff, being able to anticipate underachievement enables us to take proactive actions to achieve the contractual level of traffic. This way, all parties can keep benefiting from the lowest costs or avoid bill-shocks.

Finally, ahead of contract negotiations or renegotiations with hardware or software providers, AI can identify anomalies or patterns that are useful to negotiate the most adapted contract. For example, when using cloud services, these insights will help select the most appropriate pricing model, Pay As You Go or resource reservation.

Best practices for using AI in network management

Big benefits are available using AI in network management, but it’s worth remembering a couple of best practices to get the most from it. A fundamental prerequisite to using AI is cleansing data, removing duplicates, and ensuring that data is always refreshed regularly. Further, it’s also worth remembering that integrating AI into legacy IT can be less than straightforward. Algorithms don’t always integrate natively with existing IT systems, and so some IT architecture planning can be required for this integration.

A powerful tool that brings big benefits

The future of network management lies in harnessing the power of AI and using it to your advantage. Our customers need their network infrastructure to work as flawlessly as possible as much of the time as possible.

The above use case examples demonstrate Orange Wholesale International’ commitment to leveraging AI’s power in our networks to make them more efficient, more agile, and to enhance our customers’ experience and satisfaction.

However, AI isn’t being used simply to replace IT teams, but to augment them. We see it playing a key role in binding disparate systems and processes together so that IT teams are better equipped to make sense of the constant flow of data and enhance network management. But this is just the beginning: we’re investigating many more use cases to keep driving forward and making full use of the full power of AI.