Published On 7/6/2026
|
Last update: 14:31 (Mecca time)
In an unprecedented move, the Argentine government proposed a project aimed at creating a legal framework that would allow the emergence of what it called “Non-Human Corporations,” entities that could be managed by artificial intelligence systems and digital agents instead of traditional human management.
If the project is approved, Argentina could become the first country to establish a legal basis for this type of company, while ultimately maintaining human oversight and responsibility.
Read also
list of 2 itemsend of list
Although the legal and ethical controversy has captured the attention of the media, the real value of the initiative lies in its technical structure, as it reflects the global trend towards what is known as agentic artificial intelligence (Agentic AI), where the role of the system is not limited to generating texts or answering questions, but rather extends to planning, decision-making, carrying out tasks, and interacting with digital systems semi-autonomously.
From intelligent assistant to digital manager
Traditional companies rely on enterprise resource planning (ERP) systems, accounting and customer management software, but they always need employees to analyze data and make decisions. In the new model, artificial intelligence agents become the primary operational layer within the organization.
This model is based on the integration of several technologies, most notably large linguistic models (LLMs) to understand information, communicate, and make initial decisions, machine learning to analyze data and predict trends, multi-agent systems that distribute tasks among several specialized agents, and application programming interfaces (APIs) to link artificial intelligence to accounting, inventory, human resources, and banking systems.
Instead of having a single manager, the company may include a group of agents, one of whom is responsible for purchasing, another for customer service, and a third for risk management, while a central agent coordinates decisions between them.
How does the company make its decisions?
Technically, the decision-making process goes through several sequential stages, which are:
- Collect data from internal databases, sales systems, inventory movement, financial markets, and even news.
- Analyze data using machine learning models capable of detecting patterns and predicting outcomes.
- Propose a decision via linguistic models that generate different alternatives and evaluate the merits of each.
- Automatically implement the decision through integration with digital systems, such as issuing purchase orders, adjusting prices, or redistributing resources.
- Monitor results and learn from them to improve future decisions.
This method is known as “closed-loop decision making”, where the system evaluates the results of each decision and constantly adjusts its behavior based on new data.
Required technical infrastructure
An AI-driven company cannot run without a robust infrastructure that includes high-performance cloud computing to host and run models, real-time databases to process data as it comes in, a unified data architecture that connects all parts of the company, and monitoring systems that allow humans to track the performance of agents and intervene when needed.
While studies on the readiness of Latin American countries for artificial intelligence indicate that the success of such initiatives depends to a large extent on the availability of digital infrastructure, energy, communications, and technical personnel, and not on legislation alone.
Multi-agent systems…the backbone of the smart company
The concept of multi-agent systems is one of the most important technologies that makes the idea of “non-human companies” applicable. Instead of relying on a single artificial intelligence model, responsibilities are distributed among a group of specialized agents, so that each agent has a specific role and clear powers, and communicates with other agents to coordinate work and make decisions.
For example, a market analysis agent might monitor price changes and customer behavior, while another agent manages inventory and supply chain, a third agent manages finances, and a fourth agent serves and interacts with customers via digital channels. A “coordinating agent” collects the outputs of these agents and turns them into an integrated operational plan.
Research published on the arXiv platform indicates that this model represents one of the most prominent trends in the development of institutional artificial intelligence, because it allows the distribution of tasks and improves efficiency and flexibility compared to relying on a single multi-functional model.
Companies such as Microsoft, Google Cloud, and Salesforce have also begun developing platforms that allow the building of artificial intelligence agents that collaborate to accomplish business tasks within organizations, which reflects the sector’s transition from the “independent agent” stage.
Integration with institutional systems
AI cannot run a company in isolation from existing digital systems, so the success of this model depends on integration with the enterprise infrastructure via application programming interfaces (APIs).
When a decision is made to purchase raw materials, an AI agent can communicate directly with the ERP system, then review inventory, submit a purchase order, update financial records, and notify the logistics department, all within seconds without direct human intervention.
This level of integration is one of the most important reasons why major organizations are interested in agentic artificial intelligence, as Gartner reports indicate that organizations are moving towards using agents capable of carrying out operations, and not just providing recommendations, with this trend expected to expand in the coming years.
Cybersecurity…the biggest challenge
The more powers AI has within a company, the higher the security risks. If the system gains access to bank accounts, customer data, or production systems, any hack or software error could lead to major losses.
Therefore, companies that use this type of system rely on a set of technical controls, including applying the principle of least privilege so that each agent gets the minimum necessary powers, using multi-factor authentication and managing digital identities, recording all decisions and operations in audit logs that are reviewable, in addition to keeping a human inside the decision circle (Human-in-the-Loop) when carrying out sensitive operations such as financial transfers or signing contracts.
Regulatory bodies, including the US National Institute of Standards and Technology (NIST) underscore AI Risk Management, that human supervision, risk management, and interpretability of decisions are key to ensuring the safe and responsible use of these systems.
Are we facing a completely independent company?
Despite great ambitions, most experts believe that the concept of a “completely AI-run company” is still a long way off. Even with the rapid development of linguistic models and intelligent agents, these systems still face challenges such as “hallucinations,” the weak ability to deal with unexpected situations, and the need for human oversight in decisions that have a legal or financial impact.
For this reason, the Argentine initiative does not grant artificial intelligence an independent legal personality, but rather proposes a model in which operational management is largely automated, with a human administrator continuing to bear ultimate legal accountability, as reported by Reuters and the Financial Times.
