Turning Complexity Into Opportunity: How AI Can Transform Service Assurance in Telecom

Here's a truth we've learned through experience: You cannot move up the autonomy scale without a solid data strategy. Full stop.

Data awareness is everything in AI and machine learning. You can have the most sophisticated algorithms and the best intentions, but without organized, accessible, high-quality data, you're building on sand.

This isn't optional. It's the foundation that must be in place before you can progress from basic automation to advanced capabilities.

Four Barriers We've Seen in the Field

Through our work with operators around the world, we've identified four significant barriers that slow or stop the journey to automation:

1. Corporate Culture

This might surprise some, but culture is often the biggest challenge. Moving from repetitive, straightforward manual work to relying on AI-driven systems requires a fundamental shift in mindset. Your teams need to trust the system to handle their daily tasks and then focus on higher-value activities. That's not a technical problem - it's a human one.

2. System Integration Issues

Too much confidence in what providers and hyperscalers can deliver - or conversely, too little confidence - creates hesitation. The vertical silos that have existed for years need to give way to integrated systems with accessible APIs and easy integration paths. As vendors, we share responsibility for creating these silos, and we need to work together to break them down.

3. Data Availability

Sometimes data availability is your biggest barrier to achieving specific goals. In service assurance, success isn't about reports, KPIs, or dashboards - it's about finding the right data at the right time to make the right decision. When that data isn't available or accessible, progress stops.

4. Privacy Regulations

The EU's AI regulation and GDPR create legitimate concerns. Operators need to balance the amount of data consumed and exposed to AI models against their obligations to protect subscriber privacy. As these regulations mature, operators need clarity on where the boundaries lie.

Here's a truth we've learned through experience: You cannot move up the autonomy scale without a solid data strategy. Full stop.

Data awareness is everything in AI and machine learning. You can have the most sophisticated algorithms and the best intentions, but without organized, accessible, high-quality data, you're building on sand.

This isn't optional. It's the foundation that must be in place before you can progress from basic automation to advanced capabilities.

Four Barriers We've Seen in the Field

Through our work with operators around the world, we've identified four significant barriers that slow or stop the journey to automation:

1. Corporate Culture

This might surprise some, but culture is often the biggest challenge. Moving from repetitive, straightforward manual work to relying on AI-driven systems requires a fundamental shift in mindset. Your teams need to trust the system to handle their daily tasks and then focus on higher-value activities. That's not a technical problem - it's a human one.

2. System Integration Issues

Too much confidence in what providers and hyperscalers can deliver - or conversely, too little confidence - creates hesitation. The vertical silos that have existed for years need to give way to integrated systems with accessible APIs and easy integration paths. As vendors, we share responsibility for creating these silos, and we need to work together to break them down.

3. Data Availability

Sometimes data availability is your biggest barrier to achieving specific goals. In service assurance, success isn't about reports, KPIs, or dashboards - it's about finding the right data at the right time to make the right decision. When that data isn't available or accessible, progress stops.

4. Privacy Regulations

The EU's AI regulation and GDPR create legitimate concerns. Operators need to balance the amount of data consumed and exposed to AI models against their obligations to protect subscriber privacy. As these regulations mature, operators need clarity on where the boundaries lie.

Here's something we spend enormous amounts of time on that rarely gets discussed: data unification.

Every data source comes in different schemas and formats. Fault management data, performance management data, customer experience data, customer care data - they all arrive in different shapes. You cannot imagine how much effort it takes just to unify these sources to enable a single use case.

If you want to improve customer experience management, you need to combine all these disparate data types. That unification work is invisible to end users but absolutely critical to success.

Four Layers That Determine Your Progress

When we discuss automation journeys with customers, we focus on four foundational layers that determine how fast you can progress:

• Tools and Applications: Can you afford multiple tools to troubleshoot the same problem, or do you need to consolidate? As suppliers, we need to provide accessible APIs and easy integration. The industry needs to move beyond vertical silos.
• Foundational Functionality: Should you use open-source large language models or rely on publicly available LLMs? This choice between control and convenience has real implications.Privacy requirements may limit your use of public LLMs in certain scenarios, making this a critical strategic decision.
• Infrastructure: Public cloud, hybrid, or private? We've seen customers move completely to the cloud only to discover that migration costs don't equal benefits. The balanced approach - some applications in the cloud, some on-premises for operational efficiency - is becoming more common.
• Data Storage and Movement: Data lake, data warehouse, or lakehouse? These decisions impact your AI journey significantly. But here's what's often overlooked: moving data costs money. The cost of moving data between silos can exceed storage costs. Yet you still need that data movement to train models and power use cases.

Here's something we spend enormous amounts of time on that rarely gets discussed: data unification.

Every data source comes in different schemas and formats. Fault management data, performance management data, customer experience data, customer care data - they all arrive in different shapes. You cannot imagine how much effort it takes just to unify these sources to enable a single use case.

If you want to improve customer experience management, you need to combine all these disparate data types. That unification work is invisible to end users but absolutely critical to success.

Four Layers That Determine Your Progress

When we discuss automation journeys with customers, we focus on four foundational layers that determine how fast you can progress:

• Tools and Applications: Can you afford multiple tools to troubleshoot the same problem, or do you need to consolidate? As suppliers, we need to provide accessible APIs and easy integration. The industry needs to move beyond vertical silos.
• Foundational Functionality: Should you use open-source large language models or rely on publicly available LLMs? This choice between control and convenience has real implications.Privacy requirements may limit your use of public LLMs in certain scenarios, making this a critical strategic decision.
• Infrastructure: Public cloud, hybrid, or private? We've seen customers move completely to the cloud only to discover that migration costs don't equal benefits. The balanced approach - some applications in the cloud, some on-premises for operational efficiency - is becoming more common.
• Data Storage and Movement: Data lake, data warehouse, or lakehouse? These decisions impact your AI journey significantly. But here's what's often overlooked: moving data costs money. The cost of moving data between silos can exceed storage costs. Yet you still need that data movement to train models and power use cases.