AI Agents: Automation in Business, Blurred Responsibility

AI Agents in Business: Automation Without Accountability, a Major Challenge

Companies are increasingly adopting AI agents to automate the execution of certain tasks and decisions. However, a crucial question remains: who is responsible when these agents act autonomously?

Initially, generative AI primarily served as an assistance tool. AI agents represent a significant advancement in that they no longer just advise but make decisions and execute actions autonomously.

This shift fundamentally alters the distinction between recommendation and action. It raises a question often overlooked: who should be held accountable for the decisions and actions of these agents? Is it the IT department managing the systems, the business teams setting the objectives, the data teams designing the models, or the legal services ensuring compliance? Each plays a role, but none assumes total responsibility.

The real issue here is not ownership but delegation. Companies are now allowing systems to act without clearly defining who bears the consequences.

Behind this question lies a strategic choice. The adoption of AI agents goes beyond mere technological considerations and becomes a major organizational decision that will influence whether these systems create economies of scale or further fragment operations.

Systems Executing Without Clear Supervision

For a long time, the IT architecture of companies was relatively stable. Infrastructures supported applications, which manipulated data, while users made decisions.

AI agents disrupt this balance. They constitute a new software layer capable of orchestrating actions across various systems. They interpret instructions, utilize data, and act directly through interfaces or APIs. They no longer just provide information: they produce tangible effects.

This capability transforms agents into true digital operators. Their impact directly depends on the systems they are connected to. An agent limited to an internal documentation tool remains relatively harmless. In contrast, an agent connected to a CRM, a customer database, or a billing system can influence critical operations.

For example, an agent tasked with qualifying leads can automatically redirect business opportunities. A misconfiguration does not merely result in a bug; it can directly affect revenue.

The adoption of AI agents is progressing rapidly. Approximately 35% of companies are already using them in certain functions, and 44% plan to deploy them soon. Yet, less than half of these companies report having a clear framework to govern their use. The gap between technological adoption and organizational maturity is already beginning to widen.

When systems make decisions and execute actions, the question of accountability becomes central.

Centralization vs. Autonomy: A Debate Missing the Point

As they grow, companies face a well-known tension. Should control of the agents be maintained through the IT department, or should business teams be allowed to progress more quickly with autonomy?

Centralization makes sense because it allows for controlling access to data, securing interactions with critical systems, and keeping track of automated actions. In a regulatory context reinforced by the AI Act in Europe, this requirement becomes essential.

However, excessive centralization can quickly become a hindrance, and the most relevant use cases often emerge on the ground, where processes are imperfect and pain points are well identified.

Allowing business teams to experiment freely accelerates innovation, but at the cost of a real risk: the emergence of a multitude of agents developed without coordination, connected to sensitive and sometimes invisible data for the IT department. A situation reminiscent of shadow IT, except that these systems do not just observe; they act.

But this debate is poorly framed. Centralizing or decentralizing does not answer the essential question: who is accountable for the decisions made by the systems?

A Still Vague Chain of Responsibility

Currently, responsibility is distributed without being truly assigned. Technical teams develop, business teams use, and compliance functions oversee. But no one is explicitly responsible for arbitrating between local performance and global impact. This point becomes critical when agents optimize specific indicators. An agent may improve a conversion rate while degrading the customer experience or creating operational risks.

In this context, this transformation gives rise to new roles. Some companies are beginning to recruit AI architects tasked with designing agent architectures and overseeing their deployment. Others are seeing the emergence of specialized profiles by function, within RevOps, growth, analytics, or digital teams, who design agents directly integrated into business processes.

The question remains where to position them. In the IT department, to maintain control over the systems? In the business teams, to stay close to the usage?

The Mirage of the Hybrid Model

In the face of these tensions, many companies are converging towards a hybrid model: a centralized platform combined with business autonomy. On paper, this balance seems relevant. In practice, it often masks a lack of decision regarding accountability.

Today, few organizations explicitly designate a person responsible for the decisions made by the agents. In the event of an error, accountability dilutes among technical teams, business teams, and control functions. Governance exists in diagrams but disappears in execution. An organization must define what an agent is allowed to do, within what limits, and in which cases a decision must be validated or escalated to a human. It must also be capable of supervising its actions and auditing their effects.

This model relies on a clear distinction between the technical platform and business usage. The agent infrastructure remains under the responsibility of a central team, often linked to the IT department. This team manages access to data, security mechanisms, action supervision, and audit tools.

Business teams retain responsibility for use cases and operational outcomes. They define what the agent should accomplish and assess the value generated.

However, this model only works if explicit responsibility is assigned in the event of failure. As long as a role is not clearly mandated to respond to automated decisions, the hybrid organization becomes a dilution of responsibility rather than a balance.

Responsibility to Be Explicitly Reassigned

AI agents will quickly establish themselves as a key component of digital operations. Their deployment raises less of a technological question than a governance issue.

Companies have learned to structure the responsibility of IT systems. They have yet to structure that of systems that act.

In the coming years, the advantage will go to companies capable of clearly organizing delegation and accountability. For technological autonomy does not eliminate human responsibility; it makes its absence immediately visible.