Autonomous networks are networks that, under most conditions, can self-configure, self-monitor, self-optimize, and self-heal. The emergence of Agentic AI further underscores the fast transition to more autonomous operations.

Operators are increasing investments in AI to gain a competitive edge in driving towards higher levels of autonomy.

Accelerating maturity in autonomous networks journey is a strategic challenge. There are around 800 million dollars of value to capture of combined revenue upside and operational cost reduction.

However, only 6% of CSPs report on Level 4 autonomous networks today[1].

There is a reason for that: most of the time the technological components cannot be reused to enact an autonomous networks program of change.

Achieving autonomous operations is not a trivial task, as it requires implementing these 3 constituent pillars:

Data:

• Reference Telco Data Platforms act as central hub for telecom operators to manage large-scale network and customer data.
• Data Mesh & Data Products, enabled by data mesh architecture, a shift from monolithic data lakes to federated, productized data services. You define a network quality of service data product once and reuse for many use cases related to network performance.

Analytics

• The core of autonomous operations is analytics. Analytics encompasses real time analysis, root cause detection, event correlation, pattern recognition, trend analysis, forecasting & impact, analysis, and prescriptive analysis. These capabilities enable to identify and prioritize issues, diagnose, and troubleshoot problems, optimize, and tune performance, and prevent and mitigate risks.

Automation

• Intent-Based Automation lets operators set high-level goals, which the system turns into workflows. Agentic AI Platforms and adjacent technologies such as Digital Twin, will enable multiple AI agents to coordinate complex tasks such as managing intent conflicts and closed loop assurance.

Let us start with the first, the data.

Why data modernization?

We constantly talk about data as an asset, that data is gold, but we do not realize the value of such assets. In the last ten years, we have used domain driven design effectively for our operational systems, but we have ignored the domain concepts in a data platform.

We congratulate ourselves on creating the biggest monolith ever, the big data platform as an evolution of the data warehouses and the data lakes and data oceans. This slows down the process of getting data from different sources to support the drive for GenAI.

On average, telecoms take 6 to 9 months[2] to put in production an AI-enabled application.

If we want to accelerate the journey of a telco becoming an AI-telco, it is important to adopt the following principles:

• Data domains: shifts from theoretical or aspirational reference industry frameworks to domains should match reality, aligned with business capabilities, which the business is in, this requires the definition of an ontology.
• Data products: shifts the value system from data as an asset, to data-as-a-product to share and connect internal and external users. To be feasible from both a data availability & technical standpoint. To become a source of monetization. Data products are mapped to domains – network, sales … to ensure data is engineered & provided for consumption across the organization. Data products can be reports, datasets, ML models to be offered as-a-service with federated security governance, enacting data products with infrastructure as a code.
• Self-serve platforms: shifts from collecting data in siloed data warehouses and data lakes. shifts from two sets of fragmented integrated infrastructure services. To allow data consumers to self-serve from a pool of data products, governed by standardized policies.
• Federated governance: shifts from collecting data in siloed data warehouses and data lakes. Shifts from two sets of fragmented integrated infrastructure services, towards allowing data consumers to self-serve from a pool of data products, governed by standardized policies. To distributed infrastructure architecture that enables composability and reusability.

What are data products?

A trusted, reusable, and consumable data asset that delivers curated, structured datasets enriched with business-approved metadata and domain logic. Designed to support specific business needs, data products enable efficient data discovery, decision-making, and AI-driven applications. They are:

• Prepared: Data is thoroughly cleaned, transformed, and validated to ensure the highest quality standards.
• Findable & Understandable: Comprehensive metadata enables straightforward discovery and comprehension.
• Interoperable: Seamlessly integrates with other datasets to provide comprehensive insights.
• Shareable & Accessible: Securely packaged to support governed and compliant distribution.
• Reusable: Designed to be modular and composable for diverse applications.

To enhance network performance, especially for anomaly detection, it is essential to establish a dedicated network quality of service data product. This data product, when integrated with a customer 360 data product, demonstrates significant value by supporting scenarios such as churn prevention. The ability to reuse these data products across different use cases highlights the advantages of a model-driven approach, emphasizing modularity, reusability, and composability in solution design.

Moving towards distributed domain driven architecture will be the biggest opportunity

The adoption of data products represents a foundational step toward reducing the time required to deliver solutions. By leveraging standardized and reusable data assets, organizations can eliminate prolonged development timelines—such as the traditional six-month cycle to reach production—thereby accelerating time to value and enabling more agile responses to business needs.

[1] Navigating autonomous networks, IBM, https://www.ibm.com/thought-leadership/institute-business-value/en-us/report/autonomous-networks

Accelerating the adoption of telco AI to deliver autonomous networks, Analysys Mason 2023, https://www.analysysmason.com/research/content/perspectives/telco-ai-autonomous-networks-apr2023-rma14/

[2] Building an AI strategy: telcos put the foundations in place, TM FORUM, 2024- https://inform.tmforum.org/research-and-analysis/reports/building-an-ai-strategy-telcos-put-the-foundations-in-place