Understand the job
Purpose, instructions and context shape what an agent is trying to achieve.
TECHNICAL PAGES
AI agents are more than chatbots or prompts. This guide explains the building blocks that help an agent do useful work inside a real organisation.
An AI agent combines an objective, instructions, reasoning, context, knowledge, tools, integrations, state, actions, controls and human involvement. Different agents use different combinations of those elements.
A simple conversational agent may use only a few. A sophisticated operational agent may work with several information sources, systems, workflows and people. There is no single universal architecture; these pages explain the major building blocks so business leaders can ask better questions about the work an agent is intended to support.
Purpose, instructions and context shape what an agent is trying to achieve.
Knowledge, tools, integrations and state determine what it can reliably do.
Human review, permissions and governance define where responsibility remains.
The most useful question is not which technology to buy. It is what outcome is needed, what information is required, what actions are permitted and where people must remain accountable. Explore practical examples in AI Agents for Compliance.
PRACTICAL LENS
Understanding agent architecture helps a firm move beyond isolated prompting and design complete operating capabilities: clear jobs, useful information, controlled actions and accountable people. It provides a foundation for prioritising AI work that can genuinely fit the firm's operating model.
PRACTICAL LENS
The same understanding becomes important if a firm chooses to develop AI-enabled services. A repeatable service needs its purpose, client information, tools, controls and human responsibilities deliberately designed — not simply a conversational interface added to an existing offering.