The goals of an agentic architecture

To move from a few agentic pilots to larger deployments, telcos must focus on developing a range of new technologies, which can be visualised as situated within an “agentic architecture”.

The architecture articulates the layers of technology and supporting processes required to build agentic systems:

Human: the architecture places the needs of users at the top of an agentic system to ensure focus on building trust and good user experience

Apps: apps interact with both humans and agents.  These may be tools internal to the telco (e.g. CRM system) or external (e.g. customer or partner systems)

Agents: deployment of simple agent systems (which are controlled from the top of a hierarchy) and, over time, multi-agent systems (where agents obtain a degree of autonomy)

Intelligence: deployment of appropriate models to power agents and other tools

Knowledge: the federation and organising of knowledge to support model decision-making with suitable context

Data: the gathering, federation, transportation and use of data for training and use in the models

Compute and storage: supporting hardware, software and cloud resources

Operations layer: processes and technology that support the lifecycle of agents, models, knowledge and data. Security, privacy, governance, compliance and risk are also considerations here.

The top-level goals of each layer are: