An Introduction to Agentic AI for Governments
The long-standing trend of 'software eating everything' describes a fundamental shift in which software-based solutions are replacing traditional processes, physical hardware, and even entire industries. The rapid advancement of AI is accelerating this trend and pushing its boundaries into areas that were previously considered to be the exclusive domain of humans. This is fundamentally altering the nature of what is 'being eaten'.
Previous rounds of digital transformation transformed industries by virtualising infrastructure (e.g. cloud and software at the core of hardware, from cars to radios), dematerialising products (e.g. music, films, and books) and processes (e.g. e-commerce and e-banking), and automating rule-based tasks (e.g. business process automation and robotic factories).
Generative AI, with its capacity to create novel and human-like content (e.g. text, images, audio, code, or synthetic data), is now ‘eating’ functions that involve complex cognition, creativity and understanding, such as content creation and communication, software development, customer engagement, data analysis, public administration, education, and research and development (R&D). In many executive, academic, public administration, and research-related roles, generative AI is already seeing widespread adoption.
AI agents take this a step further by introducing systems that can not only generate and analyse data, but also perceive, reason, and act with minimal human intervention. These so-called agentic systems can manage end-to-end processes, learn, self-optimise, and collaborate with humans and other agents (see table below for different types of agents).
Types of agents currently in use
Type | Description | Examples |
Information and Analysis | These agents specialise in gathering, processing, synthesising, and analysing information from diverse sources. Their primary function is to provide insights, answer complex queries, make comparisons, or offer decision support. | Agents performing product comparisons by browsing the web or personalised agents assisting individuals in making complex decisions, such as selecting insurance plans or educational institutions. |
Task Execution and Automation | These agents are designed to perform specific actions and automate workflows within digital (and sometimes physical) environments. They follow instructions or pursue defined goals to complete tasks. | Agents that buy goods online, book flights, conduct software engineering tasks, or provide office support by managing schedules and documents. |
Interaction and Communication | These agents focus on engaging with humans or other AI agents. Their capabilities centre on natural language understanding, dialogue management, negotiation, and facilitating collaboration. | Agents capable of making phone calls or engaging in complex interactions with other agents using specialised communication protocols to coordinate activities or make agreements. |
Systems of agents, what we refer to as agentic AI, are predicted to take over tasks that previously required significant human oversight and judgement, such as software engineering, IT operations, human resources, finance, customer service, marketing, operations, and supply chain management (see figure below).
Job function groups with the highest exposure to generative AI¹
The Agentic State
The rise of agentic AI will have a particularly profound impact on government and public services, as well as the provision of public goods such as education and healthcare.
Previous rounds of digital transformation have altered the medium by government functions, replacing paper forms and physical front offices with online forms and apps or portals, but have not altered the fundamental organisational structure or business model of government.
This time will be different. Agentic AI's software will be capable of 'eating' the core functions of government. The Agentic State is not simply the automation of existing processes or the replacement of human rules with AI agents, but marks a shift in the nature of government and public administration on par with the invention of the bureaucratic state in the 19th century (which was itself a reflection of the technological and organisational paradigms of the Industrial Revolution).
The Agentic State will be driven by outcomes, not process; customised and personalised, not consistent and formulaic; real-time, not predictable; and will orchestrate public goods without acting itself. The associated changes may be profound, presenting new risks, but they may well be inevitable if government and public administration are to maintain their fundamental role in delivering public value and serving the public interest.