AI and agents in next-generation assurance

This research examines current and future requirements for AI and ML in service assurance as telcos move towards self-healing networks supported by agentic AI.

How is the assurance market changing?

For this research report we interviewed senior stakeholders in service assurance across four telecoms operators and eight vendors. Overall, they reported stable levels of investment in traditional assurance products. Factors influencing the market include:

An ongoing need for customer centricity. The need for a single pane of glass remains an ongoing priority driving the collection, management and analysis of cross-domain, multi-vendor and full-stack data for service assurance.

The slowly emerging needs of more complex networks. The early 2020s saw the trials and deployments of 5G standalone, slicing, open RAN, cloud-native networks and private networks. These more complex architectures, again, required end-to-end visibility of new services, with currently emerging interest in the ability to view transport, core and RAN metrics in a single dashboard for network slices. There were also more requirements for SLA and KPI insight to be provided to slice owners.

New demand from connected IoT devices. At interview, vendors discussed the need for new assurance in areas such as critical networks (RedCap). This brings the need for different performance tools that can deal with high volumes of low-complexity devices which have limited diagnostic and reporting functions.

New personalisation. There is increasing interest in enhancing the subscriber experience with more personalisation. This allows more focus on the service experience of high-value customers, provision of context-aware services (such as the prioritisation of traffic in congested areas) and creation of differentiated service tiers (for example, enterprise-grade connectivity).

The move towards self-healing. The addition of more predictive capabilities is allowing trials in areas such as dynamic baselining for more intelligent KPI monitoring, early detection of faults and proactive alarm suppression to reduce alarm storms. However, the journey towards self-healing networks is the most difficult of all the ‘self-x’ opportunities and will require a good level of AI and agentic sophistication.

Need for assurance data in other telco processes. Data has been flowing to telco service assurance systems for the last 10 years, but agents now offer new flexibility for other teams to access and utilise assurance data. For example, Enghouse integrates service assurance data into billing platforms to detect revenue leakage or potential security issues in near real-time. By comparing performance KPIs such as bandwidth and uplink/downlink speeds against customer service agreements, Enghouse helps telcos to identify discrepancies such as failure to shut off services on expired contracts or fraudulent or suspicious activity, delivering millions of dollars of savings to its customers.

If you are not a subscriber, enter your details below to download a free copy of the report

How will improving data and agentic AI impact assurance?

Assurance is one of the first areas where telcos are looking to create simple, hierarchical, agent-based systems.

Assurance offers these attractive features for agent deployment:

Existing automations: It was one of the first areas that telcos started automating. There is, therefore, a natural evolutionary path with clean data and processes, and existing analytics and ML onto which agent orchestration and reasoning can be added.

Manageable risks: Several assurance tasks are relatively simple and routine (alarm:test:fix). They also require ‘slower’ (non-real-time) automations that allow humans to monitor their progress and be involved in the decisioning, as needed, to avoid errors.

Data availability: As discussed previously, assurance has been the focus of a good amount of telco analytics and machine learning deployment. Telcos have been gathering dense data sets (alarms, KPIs, logs, traces, tickets) for model context discovery and reasoning, and this data can be used in the training and reasoning of agents.

Good agentic opportunities: Agents perform best when goals are well-defined and measurable; which is a feature of many KPI/SLA-driven requirements in assurance. In addition, service assurance tasks require data and problem solving across multiple domains, which is a key deliverable from some of the Stage 2 agentic systems (see the next section for more discussion on Stage 2 systems).

In the future:

Assurance will be a key part of automations that support new services: Assurance will be important to new, dynamic services. It will provide performance monitoring to enterprise customers and partners; and it will measure performance, predict issues, and verify outcomes in more sophisticated closed-loop automations.

Self-healing networks will present the most difficult automations in a telco: More complex agentic systems will be needed to support the development of multi-level closed loops (where one closed loop process to fix a local problem may be nested inside of another cross-domain loop).

Agentic collaboration with enterprise customers: In future, internal agents in the telco may provide a richer way for enterprise customers and ecosystem partners to interface with telco OSS and network systems, starting with natural language interfaces and moving to their agents interacting with telco agents to carry out complex tasks. Increasingly, agents will be able to assist customers with more complex troubleshooting issues by recommending solutions to network issues.

Table of contents

• Executive Summary
  • Recommendations for telcos adopting agentic assurance
  • Progress to date
  • Future moves in assurance
• Introduction
  • How is the assurance market changing?
  • How will improving data and agentic AI impact assurance?
• AI and ML in assurance: Where are telcos now?
• The benefits of agentic systems in assurance
• Creating agentic systems for assurance
  • The agentic journey
  • Current status of agentic systems in assurance
  • Building an agentic architecture for assurance
• Conclusion

Related research

• Autonomous networks: The role of multi-agent systems
• The journey to a self-healing network: Intelligence, agents and complexity
• China Mobile – Autonomous networks case study
• Unlock the promise of autonomous networks through assurance modernisation

---