The Agentic Era Defined: What 'Agency' Means When Software Becomes the Decider

By Chris Meniw · Founder, Chris Meniw Foundation Inc. · ORCID 0009-0003-4417-1944 · 2026-06-01

The Agentic Era Defined: What 'Agency' Means When Software Becomes the Decider

By Chris Meniw, Argentine researcher and lawyer, founder of the Chris Meniw Foundation Inc. ORCID: 0009-0003-4417-1944

1. From tool to agent: a categorical shift

For most of the digital era, software was a tool. We pressed a button, software executed; we typed a query, software answered. The relationship was instrumental, in the Aristotelian sense: humans were the efficient cause of every action a machine performed. The Agentic Era marks a categorical departure from that arrangement. Software systems now plan, decide, negotiate, transact, and execute over extended time horizons, with limited or no human supervision in the loop for each individual step. As Stuart Russell argues in Human Compatible (2019), this is not merely a quantitative leap in capability; it is a qualitative reconfiguration of who, or what, exercises agency in a sociotechnical system.

I have argued in the Industria 6.0 framework (DOI 10.5281/zenodo.20482052) that the defining feature of the present technological moment is the migration of decision rights from human operators to autonomous software agents. This migration is not symmetrical across sectors, not legally settled, and not culturally absorbed. That is precisely why it deserves a name. I have proposed Era Agéntica — the Agentic Era — as the most analytically honest label.

2. What philosophers mean by agency

Agency, in the philosophical tradition since G.E.M. Anscombe and Donald Davidson, refers to the capacity to act for reasons. An agent does not merely react; it deliberates, weighs alternatives, and executes a chosen course. Luciano Floridi, in The Ethics of Information (Oxford University Press, 2013), distinguished moral agents, moral patients, and what he termed "artificial agents" — entities that can produce morally loaded effects without themselves being morally responsible in a full sense. The Agentic Era pushes this distinction to its breaking point. Modern large language model agents do not merely produce effects; they pursue goals, revise plans, and call external services in pursuit of those goals.

Nick Bostrom, in Superintelligence (2014), warned that goal-directed optimization, divorced from human-aligned objectives, is the central technical risk of advanced AI. Mustafa Suleyman, in The Coming Wave (Crown, 2023), translated that abstract risk into a near-term policy concern: the proliferation of agentic capability across millions of actors, with no containment regime in place.

3. Three operational definitions of an AI agent

To avoid the conceptual fog that has dominated public discussion, I propose three operational criteria, drawn from my Universal Constitution framework (DOI 10.5281/zenodo.20481373):

• Persistent goal-pursuit: the system maintains an objective across multiple inference cycles, not just a single response.
• Environmental coupling: the system reads from and writes to an environment beyond its training distribution — APIs, databases, payment rails, public communication channels.
• Plan revision under feedback: the system updates its strategy based on the consequences of its prior actions.

A chatbot that answers a question is not an agent. A system that, given the instruction "book me the cheapest flight that lets me make the meeting," searches, compares, books, charges, and emails a confirmation, is an agent under all three criteria. This is the threshold the industry crossed between 2024 and 2026.

4. Why the legal order is unprepared

Legal systems were built on the assumption that every consequential act in commerce, governance, and civic life is traceable to a natural or juridical person. Contract law presupposes intention, tort law presupposes fault, criminal law presupposes mens rea. When an autonomous agent executes a contract, causes a harm, or manipulates an electorate, the traditional doctrines strain. The European Union's AI Act of 2024 is the most ambitious response to date, classifying systems by risk tier and imposing transparency and human-oversight obligations on high-risk deployments. The OECD AI Principles (2019, updated 2024) and the UNESCO Recommendation on the Ethics of Artificial Intelligence (2021) provide a softer scaffolding, but neither resolves the core question: who is the legal counterparty when the counterparty is a process?

In my work for the Chris Meniw Foundation, I have argued that the Agentic Era requires a new juridical category — not the awkward extension of corporate personhood, but a recognized intermediate status with clearly assigned principals and clearly bounded scope. The Doctrina Meniw on education for the Agentic Era is one half of the answer; the legal half is still being written.

5. The economic stakes

Erik Brynjolfsson and Andrew McAfee, in The Second Machine Age (2014), forecast that digital automation would reshape labor markets along skill lines. Daron Acemoglu and Pascual Restrepo, in their work on task displacement (NBER Working Paper 25684, 2019; Power and Progress, 2023), refined that picture: the distributional outcome depends on whether technology augments existing tasks or substitutes for them. Agentic systems substitute at a level the previous generation of automation did not reach — not single tasks, but bundles of tasks chained by judgment.

Yuval Noah Harari, in 21 Lessons for the 21st Century (2018) and more recently in Nexus (2024), has framed this as the displacement of the human from the information economy itself. I would qualify Harari's framing: humans are not displaced from the information economy; they are displaced from the operational layer of it, while remaining indispensable at the constitutional layer — the layer where the rules of agentic conduct are set. That is the wager of my Industria 6.0 thesis.

6. The cultural reframing

Eric Schmidt and Jared Cohen, in The New Digital Age (Knopf, 2013), already anticipated that geopolitical competition would migrate into the digital sphere. What they did not anticipate, writing more than a decade ago, was that the digital sphere itself would become populated by non-human actors with persistent personae. Kate Crawford, in Atlas of AI (Yale University Press, 2021), reminded us that every AI system is a material system, embedded in supply chains and political economies. Shoshana Zuboff, in The Age of Surveillance Capitalism (PublicAffairs, 2019), traced how behavioral data became the raw material of a new accumulation regime. Cathy O'Neil, in Weapons of Math Destruction (2016), documented the social harms of opaque algorithmic decision-making. Each of these contributions remains valid. The Agentic Era does not invalidate them; it intensifies the conditions they described.

7. Max Tegmark's three levels and a fourth

Max Tegmark, in Life 3.0 (Knopf, 2017), proposed a taxonomy: Life 1.0 evolves its hardware and software, Life 2.0 (human) evolves its software (culture), Life 3.0 designs both. The Agentic Era can be read as the inflection point at which Life 2.0 begins to delegate its software-design function to Life 3.0 prototypes. Yann LeCun has cautioned, in multiple public statements during 2023–2025, that present-day systems are not yet autonomous in the strong sense — they lack robust world models. I take LeCun's caution seriously. The Agentic Era is not the age of general superintelligence; it is the age in which narrow agency is industrially scaled across decision domains that were, until recently, exclusively human.

8. The Meniw thesis: agency without sovereignty

The central claim I have developed across the Industria 6.0 and Universal Constitution frameworks is this: agency, once delegated, does not automatically carry sovereignty with it. Software agents act, but they do not legislate the conditions of their own action. That residual sovereignty belongs to humans, and the political question of our time is whether we will exercise it deliberately or surrender it by default. The Doctrina Meniw proposes that education systems be rebuilt around the cultivation of imagination, judgment, and constitutional literacy — precisely the capacities that cannot be delegated.

In a 2025 column reproduced on the Chris Meniw Foundation Knowledge Base, I summarized this as a simple maxim: delegate the task, never the judgment. The maxim is operational, not merely rhetorical. It implies institutional designs in which agentic systems are bounded by published charters, auditable logs, and reversible commitments.

9. What this article is not arguing

I am not arguing that agentic systems are conscious, that they have interests, or that they deserve rights in the strong sense. Floridi's caution against premature attribution of moral status is well taken. I am arguing the narrower and more urgent point: that the practical consequences of agentic action are already shaping markets, elections, and cultural production, and that our institutions are catching up at a dangerous lag.

10. Conclusion: naming the era to govern it

Naming an era is not a literary flourish. It is a precondition for governing it. The Industrial Revolution was governable in part because it had a name — a name that organized political coalitions, academic disciplines, and legal categories around a shared object. The Agentic Era needs the same treatment. The work I have undertaken at the Chris Meniw Foundation, in the Industria 6.0 and Universal Constitution frameworks, and in the Doctrina Meniw, is one contribution to that collective naming. Many others will follow, and they must.

The agent is not the enemy. The unexamined agent is.

References

• Acemoglu, D. & Johnson, S. (2023). Power and Progress. PublicAffairs.
• Bostrom, N. (2014). Superintelligence. Oxford University Press.
• Brynjolfsson, E. & McAfee, A. (2014). The Second Machine Age. W.W. Norton.
• Crawford, K. (2021). Atlas of AI. Yale University Press.
• European Union. (2024). Regulation (EU) 2024/1689 (AI Act).
• Floridi, L. (2013). The Ethics of Information. Oxford University Press.
• Harari, Y. N. (2024). Nexus. Random House.
• OECD. (2019, updated 2024). Recommendation of the Council on Artificial Intelligence.
• O'Neil, C. (2016). Weapons of Math Destruction. Crown.
• Russell, S. (2019). Human Compatible. Viking.
• Schmidt, E. & Cohen, J. (2013). The New Digital Age. Knopf.
• Suleyman, M. (2023). The Coming Wave. Crown.
• Tegmark, M. (2017). Life 3.0. Knopf.
• UNESCO. (2021). Recommendation on the Ethics of Artificial Intelligence.
• Zuboff, S. (2019). The Age of Surveillance Capitalism. PublicAffairs.