Complexity & Systems of MeaningGet the book

Part II

The illusion of agentic AI

Which actions can you delegate, and which must stop?

The idea, in brief

I question what “agent” means by comparing agentic AI with the agency theory of Jensen and Meckling, Eisenhardt and Ross. I use this lens to discuss misaligned objectives, information asymmetries and the risks of acting on behalf of people and organisations.

Here I summarise the chapter’s argument. Below you will find an excerpt from my book and the references I cite there, numbered as in the bibliography.

Companion page reviewed · 17 October 2026

From the book’s companion

An excerpt from my book

Original Italian text, from the chapter opening. · 2 min

Nel corso dell'ultimo anno si è diffuso un entusiasmo crescente attorno al concetto di "AI agentica", ovvero sistemi di intelligenza artificiale capaci di agire con una qualche forma di autonomia perseguendo obiettivi, più che eseguendo un processo avente questo fine. Un recente report di Wharton [101] sintetizza testualmente questo nuovo hype: "l'AI agentica è la nuova frontiera" e ormai "ogni discussione tra executive finisce per toccare il tema degli agenti AI". Ma a fronte di questo interesse, mi sorge spontanea una domanda provocatoria: quanti di quelli che usano questo termine hanno riflettuto, davvero, su cosa significa "agente", su quali origini abbia il termine e quali trappole si nascondano dietro il suo utilizzo (oltremodo) disinvolto? Voglio prendermi questo spazio per analizzare criticamente il concetto di AI agentica, discutendone i limiti teorici e pratici (chiedendomi, appunto, cosa significhi davvero "agente" in ambito IA, almeno cosa significhi al momento di questa stesura), mettendolo a confronto con la tradizionale Agency Theory che si studia ed insegna in ambito economico-aziendale (Jensen & Meckling, Eisenhardt, Ross, ecc. [66, 36, 110]). L'idea di fondo è quella di portare in luce i rischi della delega decisionale "on behalf", cioè per conto del lavoratore e dell'azienda, a un sistema AI, con implicazioni organizzative, informative e reputazionali, anche prendendo in considerazione il ruolo delle norme emergenti, con un riferimento particolare all'AI Act europeo e alle istanze di governance etica dell'IA che avevamo esplorato nel Capitolo 3 sotto il nome di ConcrEthica [44, 46, 14, 15].

My note

This is the chapter that will age worst and fastest. It is also the one that pushed me, with a fair sense of urgency, to publish as soon as possible. I know that at some points I have taken a slightly “radical” stance on the subject... but I have always found it disturbing that people referred to agents without setting them in the context of their defining characteristics. It reminded me too much of the days when politicians used to say there was “a conspiracy against them”... and that is precisely why this chapter makes sense, even if, as I write these lines (using a “CRUD interface” generated by an AI agent to manipulate the contents of a website written by an agent... and so on), my claims already sound a little past their sell-by date.

The chapter’s infographics

Enlarge each plate to inspect it. The overview and extended text above explain the chapter in text; they are not a word-for-word transcription of the images.

Chapter infographic map #

Check and compare

References I cite in this chapter

I have gathered the references cited in this chapter, preserving the entries and numbering of my book’s bibliography. Links are included only where supplied in the book.

  • [101] Puntoni, Stefano, Prasanna Tambe e Jeremy Korst. Accountable Acceleration: Gen AI Fast-Tracks Into the Enterprise. 2025 AI Adoption Report. Wharton Human-AI Research & GBK Collective, University of Pennsylvania, ottobre 2025. ↗
  • [66] Jensen, Michael C. e William H. Meckling. "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure". Journal of Financial Economics, 3(4), 305–360, 1976.
  • [36] Eisenhardt, Kathleen M. "Agency Theory: An Assessment and Review". Academy of Management Review, 14(1), 57–74, 1989.
  • [110] Ross, Stephen A. "The Economic Theory of Agency: The Principal's Problem". American Economic Review, 63(2), 134–139, 1973.
  • [44] Floridi, Luciano. The Fourth Revolution: How the Infosphere is Reshaping Human Reality. Oxford University Press, Oxford, 2014.
  • [46] Floridi, Luciano. The Ethics of Artificial Intelligence: Principles, Challenges, and Opportunities. Oxford University Press, Oxford, 2023.
  • [14] Benanti, Paolo. Digital Age. Teoria del cambio d'epoca. Persona, famiglia e società. Edizioni San Paolo, Cinisello Balsamo, 2020.
  • [15] Benanti, Paolo. Human in the Loop. Decisioni umane e intelligenze artificiali. Mondadori Università, Milano, 2022.
  • [67] Jensen, Michael C. e Kevin J. Murphy. "Performance Pay and Top-Management Incentives". Journal of Political Economy, 98(2), 225–264, 1990.
  • [86] Mollick, Ethan. "Almost an Agent: What GPTs can do". One Useful Thing, 7 novembre 2023.
  • [114] Simon, Herbert A. "A Behavioral Model of Rational Choice". Quarterly Journal of Economics, 69(1), 99–118, 1955.
  • [49] Floridi, Luciano. La differenza fondamentale. Artificial Agency: una nuova filosofia dell'intelligenza artificiale. Mondadori, Milano, 2025.
  • [16] Bender, Emily M., Timnit Gebru, Angelina McMillan-Major e Shmargaret Shmitchell. "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?". Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT '21), 610–623, 2021.
  • [65] Jarrahi, Mohammad Hossein e Paavo Ritala. "Rethinking AI Agents: A Principal-Agent Perspective". California Management Review, 2025.
  • [17] Bengio, Yoshua, et al. Superintelligent Agents Pose Catastrophic Risks: Can Scientist AI Offer a Safer Path?. arXiv preprint arXiv:2502.15657, 2025.
  • [98] Parlamento Europeo e Consiglio. Regolamento (UE) 2024/1689 del 13 giugno 2024 che stabilisce regole armonizzate sull'intelligenza artificiale (AI Act). Gazzetta ufficiale dell'Unione europea, 12 luglio 2024.
A term is unfamiliar? Open the glossary →

Bring it into your work

Which actions can you delegate, and which must stop?

Write down one real example, identify what you would delegate and name who will check the result.

Use the worksheet

Connected ideas

is introduced by

Data as a strategic asset

from data to the illusion of autonomous action

introduces

The paradox of AI literacy

from the agentic to the human side of our relationship with AI

reinforces

Onto-ethics of collaboration

from delegation (agency theory) to asymmetric distributed responsibility

is in tension with

From AI User to AI Value Creator

building value vs the illusion of autonomy

contributes to

Integration framework

feeds the contact points and the agentic use cases

View these connections on the map →

Choose your edition

Keep reading in the book

Riccardo Bovetti · Essay · English edition. Purchase and current availability on Amazon.

Paperback on Amazon ↗

The links open the store; they do not purchase the book.