The framework also emphasizes distributed intelligence. By minimizing data movement and enabling localized execution, it supports scalability while addressing governance constraints. At the same time, it introduces an intent-driven operational model, where workflows are dynamically determined based on objectives rather than predefined sequences.

Another key aspect is the integration of deterministic and agentic intelligence. Deterministic models provide statistically grounded insights, while agentic components coordinate investigation, escalation, and execution. This separation ensures both analytical accuracy and operational flexibility.

Bringing the Framework to Life 

The implementation of this framework is structured across multiple layers, including Codap as the data foundation and GenAie as the agentic orchestration layer.

Codap: Data Foundation and Semantic Layer 

Codap serves as the foundational data layer, providing a structured abstraction over heterogeneous telecom data sources. It ingests and organizes data from multiple operational systems, including performance management, fault management, topology and inventory systems, and customer experience data.

Through normalization and enrichment, Codap ensures consistency across domains and vendors. Data is organized using a medallion architecture, progressing from raw ingestion to AI-ready insights. This layered approach supports traceability, reproducibility, and controlled data transformation.

Codap also introduces a data product model, where curated datasets are exposed for consumption by analytical models and agents. These datasets are enriched through an ontology layer that defines relationships between network entities, enabling contextual interpretation and cross-domain correlation.

In addition, Codap integrates a knowledge base containing domain expertise, operational rules, KPI definitions, and known failure patterns. This enhances the ability of analytical and agentic components to interpret data and align insights with operational scenarios.

While Codap provides a centralized abstraction for structured and governed data, the broader architecture minimizes reliance on full data centralization by enabling localized processing across systems.

GenAie: Agentic AI and Orchestration Layer 

GenAie represents the agentic layer of the architecture, responsible for orchestrating workflows, coordinating agents, and enabling intent-driven decision-making across domains and systems.

At the center of GenAie is the Supervised Agent, which acts as an orchestration entity capable of responding to multiple types of triggers, including analytical signals, user-driven interactions, and external system requests. Based on these inputs, the Supervised Agent constructs workflows, selects relevant agents, and manages execution across systems.

Unlike traditional rule-based orchestration, this model is adaptive. Workflows are not predefined but dynamically constructed based on context, allowing the system to respond effectively to complex, multi-domain scenarios.

GenAie includes domain-specific agents such as Performance Management, Fault Management, and Trouble Ticketing agents. Each agent performs specialized tasks, including KPI correlation, alarm validation, and workflow integration. These agents operate within a task-oriented execution model, where workflows are decomposed and executed through coordinated interactions.