Developing a Strategy for Creating Autonomy in Service Assurance

How CSPs Can Build Autonomy in Service Assurance

It will be several years before CSPs can demonstrate self-healing and self-optimization across significant portions of their networks.

Driving their deployment is the need for more autonomous and dynamic systems that better support non-traditional products (such as slicing and IoT) and handle the high data volumes of 5G and 6G networks.

There are also potential savings from better supporting large teams and from creating new AI-enabled customer experiences. These enhance the day-to-day work of operational and customer-facing teams, as well as the experience across digital customer channels.

Given that most CSPs have broad agreement among their senior teams that autonomy is a key network goal, they need to determine how to generate financial upside from these opportunities. This requires a strategy that supports their journey through a diverse set of new and immature technologies, addresses the current lack of skill sets within the CSP, while demonstrating a proven return on investment.  A significant task!

Six Strategic Areas for Early-Stage Autonomy

Creating an early-days strategy requires decision-making in six areas simultaneously:

1. a vision for long-term internal and external requirements and the future agentic architecture to support them.  As always, this isn’t easy to create, but a simple vision (perhaps just one or two slides) that can be revised as understanding improves over time will inform the multiple decisions that need to be made. Helping to avoid technical debt and redundant solutions, which remain a real threat in this complex environment
2. development of a team capable of making decisions in this complex/multi-dimensional area
3. short-term focus should be on the creation of a solid foundation for this vision, particularly in the knowledge and data layer.  As well as developing the first simple agentic hierarchies, allowing the company to build human skill sets and a platform to support agents across its operations
4. the current hype around agentic systems and the immaturity of related technologies require pragmatic business cases that acknowledge the need for significant early-days spend to support the long-term vision, which might limit ROI on the first projects
5. additionally, the hype often obscures the wider range of AI, ML, simpler automations, and analytics that are already available within the CSP. When planning new autonomy, it is worth asking: What is the simplest way to solve this problem, do I already have capabilities that can be used?
6. a multi-vendor, multi-protocol, real-time service assurance solution will be required to act as a source of data and knowledge for both agents and tools in the system.  It will support not only agent hierarchies in assurance but also planning, network orchestration, marketing, sales, and operations that are complementary to a telecom’s AI strategy. It would be a great idea to feed all assurance information into an AI/ML engine for analysis to determine whether it supports the company’s strategic use cases.

See Focus areas for agentic architecture for a more detailed list of actions.

Implications for Building Self-Healing Networks
What does this mean for building out a self-healing network?

• CSPs need to create a range of simple agentic hierarchies to build understanding/skills in this next stage of assurance automation – working out what works well and building on it. Research uncovered pockets of current activity in domain-specific issue detection, root cause analysis, and agents designed to “enrich” these models with additional data
• A long-term vision that sets out some of the end-goals for assurance should include thinking around questions such as:
  • Which non-traditional products are likely to be part of our future? Are we likely to have many types of slices, our own cloud services, a full range of IoT services, and capture a slice of the AI market (where edge will become more important)?
  • In assurance, will there be a need for more complex multi-agent systems – and what will they be doing?
  • If the long-term plan includes a wide range of different products, are the plans we have today for data and knowledge architecture going to support the more complex agent systems needed, including the potential need for sophisticated agents at the edge?
• The next question should be, where are the major barriers to this vision? In assurance, the lack of ROI in building enough real-time data for some of the faster automations is a real question today.  The team creating the long-term vision should consider a gap analysis, which highlights:
  • Potential for large capital and operational expenditure
  • The financial advantage (capital or operational expenditure decrease and any revenue uplift) that will not be available if these expenditures look considerable
  • The possible workarounds (for example, use of new intelligence and tools such as synthetic data).