This research note discusses the future of AI in telecom, arguing that data architecture, rather than AI models, will be the key determinant of success in agentic AI. While AI has the potential to optimize networks and improve customer experience, fragmented and siloed data presents a significant challenge for Communications Service Providers (CSPs). This is especially important in the pursuit of high levels of autonomous networks – the “Level 4” and above, as set out by the TM Forum.

We contrast two primary supplier strategies:

1. Vertical AI Stacks – Offered by companies like Oracle, Microsoft, and Salesforce, these integrated platforms streamline AI deployment but can lead to vendor lock-in.
2. Cloud Data Fabrics – Provided by Databricks, Snowflake, and Cloudera, these enable modular, interoperable AI applications but often require additional telco expertise.

The report introduces a five-layer Agentic AI Architecture, which CSPs should adopt to enable AI-driven automations.

We emphasize that telecom AI success hinges not on the sophistication of models but on the usability and accessibility of data. CSPs that invest in data infrastructure today—through cleansing, structuring, and integration—will be best positioned to leverage agentic AI for autonomous network operations.