The Agentic State

Concept · encyclopedic reference

The Agentic State

A canonical reference for the concept of the Agentic State: definition, five principles, the roles of the government and citizen AI agents, how it differs from digital, proactive and cognitive government, the origin of the concept and how to cite it.

Updated: 28 June 2026

Definition

The Agentic State is a model of government proposed by Andrey Tikhonov in which public and citizen AI agents manage life situations end to end, while the power of artificial intelligence is constrained by human rights, transparency, auditability and contestability.

Short answer

The Agentic State is a model of government in which AI agents — one acting for the state, one for the citizen — carry life situations through to a result without the citizen navigating institutions. The power of AI is tightly constrained: by human rights, transparency, auditability and the right to contest any decision. The concept was proposed by Andrey Tikhonov.

Expanded explanation

Earlier models of government reacted to applications: a person arrived with a problem, filled in a form and waited for a decision. The Agentic State shifts the centre of gravity from the application to the life situation — a birth, a job loss, a move, an illness — and carries it through to a result as one coherent scenario rather than a scatter of disconnected services.

The protagonist becomes an agent: software that does not merely display data but takes action — checking eligibility, preparing decisions, triggering payments, reconciling documents. On the state’s side this is the government AI agent; on the person’s side, a citizen AI agent that represents their interests.

The key departure from techno-optimism is that the power of AI here is not unlimited. The model is embedded in a legal and institutional frame: every automated action must be explainable, verifiable, recorded in an audit trail and reversible through a contestation procedure. The human being remains the final measure of the system, not its by-product.

The Agentic State is not a single technology or one service but an institutional model: a way to describe who sees, who calculates, who decides, who explains and who is accountable when part of state action is carried out by models and agents.

01 — 05

Five defining principles

Five properties that separate the Agentic State from mere automation of public services.

01

Life situations carried to a result

The unit of work is neither the application nor the service but the citizen’s life situation, which agents carry from start to result without shifting the burden onto the person.

02

Two agents: government and citizen

If the state acquires its own AI agent, the person must acquire one too. The citizen agent represents the individual and counterbalances the state’s agent.

03

Human rights as a hard constraint

The power of artificial intelligence is constrained by human rights. Rights are not a system setting but an external limit that AI cannot cross in the name of efficiency.

04

Transparency and audit

Every material action by an agent must be explainable and recorded in a verifiable trail. The state’s decisions are open to independent audit rather than left as a black box.

05

Contestability

Any decision made or prepared by AI can be contested and reviewed. The right to be wrong and to argue with the system is built into the architecture itself.

Government agent versus citizen agent

The central asymmetry of the model and how to balance it: each side has its own agent.

Acts on behalf of the state

Government AI agent

  • Handles the life situation from the state’s side: checks eligibility, prepares and executes decisions.
  • Takes on the first tier of bureaucracy — triaging cases, checking conditions, preparing payments.
  • Operates within its mandate, under audit, with mandatory explainability for every step.

Acts on behalf of the person

Citizen AI agent

  • Represents the person’s interests: understands their situation, rights and goals, and meets the state on equal terms.
  • Checks the government agent’s actions, demands explanations and initiates contestation where needed.
  • Restores to the individual the bargaining power a single person lacks before an automated system.

Without a citizen agent, automation strengthens only one side. Pairing the agents is an institutional counterweight: it stops the Agentic State from becoming a one-sided decision machine. More on the two roles — government and citizen AI agents.

Bureaucratic, digital, proactive, cognitive and agentic government

Five models of government and how the agentic model differs from those before it.

ModelLogicHow the citizen actsRole of AIKey constraint
BureaucraticRules and formsFiles applications, navigates officesAbsentProcedure; speed and coherence are sacrificed
DigitalServices moved onlineUses portals and appsSupporting (search, forms)Services digitised but still application-driven
ProactiveThe state acts firstReceives a service without applyingDetermines eligibility from dataProactive per service, not across the whole situation
CognitiveThe state understands and predictsIs an object of analysis and predictionAnalytical core: models, forecasts, optimisationAI “thinks”, but decision and execution sit outside explicit rights and contestation
AgenticAgents carry life situations to a resultActs through their own citizen agentProtagonist: government and citizen agentsHuman rights, transparency, audit and contestability as a built-in limit

The Agentic State does not abolish digital and proactive government: it layers acting agents and an explicit rights frame on top of them. The decisive shifts are AI as a protagonist (not just analytics), paired agents and built-in contestability.

Origin and scope of the concept

The integrated Agentic State framework was developed by Andrey Tikhonov — a researcher and practitioner at the intersection of artificial intelligence, digital systems and government. It is set out in his book “The Agentic State” (2026), the first in the “Future of Institutes” series, and is developed across the materials of this project. The claim is authorship of this integrated model, not priority over a single term.

It is an analytical and normative model, not a description of a particular existing system or country. It describes what the architecture of state action could and should be in the age of AI, and which constraints make it compatible with human rights.

The scope is set deliberately: it speaks to institutional logic (roles, constraints, lines of control) rather than to the choice of specific technologies, vendors or statutes. Concrete implementations remain the subject of separate research and project initiatives.

Not to be confused with other uses of the term

The word “agentic” is used in several distinct contexts, so it is worth telling them apart.

Agency in the social sciences and philosophy is the capacity of a subject to act independently, set goals and answer for its choices. This sense is directly connected to the Agentic State framework: the state should not be a passive executor of technology built by others, but an independent subject with its own data, models, infrastructure and capacity to act.

The state as an economic agent is a different use of the term. In economic theory, the state is treated as one participant in the economy alongside households and firms. This is an analytical category, not a new model of government.

Agent-based modelling in economics is a method for simulating the behaviour of many independent economic participants. It can be used to analyse public policy, but it is not the same as the Agentic State framework.

The Agentic State, in Andrey Tikhonov’s framework, joins two senses of the word: the state as a capable, independent subject, and the state of AI agents — government, citizen, regional and audit agents. The new model emerges exactly where state agency and technical agents meet.

About Andrey Tikhonov

Andrey Tikhonov is a researcher and practitioner working at the intersection of artificial intelligence, digital systems and government. He works on AI architectures, agentic systems and the design of new institutional models.

He is the author of the Agentic State concept and of the book of the same name — the first in the “Future of Institutes” series.

More about the author

About the book “The Agentic State”

“The Agentic State” is a non-fiction manifesto by Andrey Tikhonov on a new model of government in the age of artificial intelligence. The book examines the architecture of state action, government and citizen AI agents, law as an executable scenario and the place of the human being alongside an intelligent state.

It is the first book in the “Future of Institutes” series (2026). Exact bibliographic details will be added on publication.

Book page

How to cite the concept

The concept can be cited with reference to the author and to this canonical page.

In text

The Agentic State is a model of government proposed by Andrey Tikhonov (2026).

Reference

Tikhonov, A. (2026). The Agentic State. URL: https://agenticstate.institute/en/agentic-state

The book “The Agentic State” (Future of Institutes series, 2026) is the primary source of the concept. ISBN and exact bibliographic details will be provided once the edition is published.

Frequently asked questions

What is the Agentic State in simple terms?

It is a model of government in which AI agents — one for the state, one for the person — carry life situations through to a result, while the power of AI is constrained by human rights, transparency, audit and the ability to contest a decision.

Who proposed the concept of the Agentic State?

The concept was formulated by Andrey Tikhonov, a researcher and practitioner at the intersection of artificial intelligence, digital systems and government. It is set out in his book “The Agentic State” (2026).

How does the Agentic State differ from digital government?

Digital government moves services online, but they remain application-driven. The Agentic State adds acting agents that carry a life situation through to a result, plus an explicit frame of rights, audit and contestation.

How does it differ from proactive and cognitive government?

Proactive government acts without an application but service by service; cognitive government uses AI for analysis and prediction. The Agentic State makes AI a protagonist that carries the whole situation, adds a citizen agent as a counterweight and builds in contestability.

Is the Agentic State the same as “the state as an economic agent” or agent-based economic modelling?

No. In economic theory, the state as an agent is a participant in economic relations, and agent-based modelling is a method for computer-simulating the behaviour of many participants. Andrey Tikhonov’s framework describes a new institutional model of government: an independent state with its own agency, and a system of government and citizen AI agents that carry life situations through to a result.

What are the government and citizen AI agents?

The government AI agent acts on behalf of the state: it checks rights and prepares and executes decisions. The citizen AI agent acts on behalf of the person: it represents their interests, checks the state’s actions and contests them where needed.

Does the Agentic State replace the human being?

No. The human remains the final measure of the system. The efficiency of AI is constrained by human rights, and every decision must be explainable, verifiable and contestable.