In the coming years, large companies will manage more non-human identities than human users. This transformation has already begun.
In the coming years, large companies will manage more non-human identities than human users. AI agents, autonomous assistants, software robots and intelligent workflows will access data, interact with applications and execute actions on behalf of employees.
This transformation has already begun.
An AI agent can today consult a customer file in a CRM, search for information in a documentary database, prepare a commercial proposal, update several systems then send a summary to a colleague. The productivity gains are real. Governance issues are just as important.
Who gave him these rights?
Who controls the data they access?
Who is responsible if confidential information is sent to the wrong recipient or if an incorrect action is performed?
For decades, companies have built their governance models around two categories of actors: users and applications. The arrival of AI agents calls this balance into question.
CIOs have spent the last decade regaining control of Shadow IT, Shadow SaaS, and application proliferation across the enterprise. However, a new phenomenon is already emerging, often outside any governance framework: AI agents.
The problem is not their existence. The problem is that most organizations still approach them as simple productivity tools while they are gradually becoming operational players in their own right.
As is often the case in digital transformation, technology advances faster than governance.
The AI agent is no longer a simple assistant
For several years, artificial intelligence remained relatively simple to manage. Employees asked an assistant, got an answer, and then remained accountable for the action.
This border is starting to disappear.
New agents no longer just respond. They execute.
They search for information across multiple applications, update data, create tickets, generate reports, or trigger workflows.
A sales agent can prepare a proposal using CRM data. A support agent can analyze a ticket, consult a documentary database and propose a resolution. An HR agent can prepare summaries from multiple internal sources.
These uses create a significant break. For the first time, software entities with a certain autonomy directly access information systems and act on them.
But our governance models were not designed for this.
For twenty years we have governed two types of identities
For decades, information systems have been based on a relatively simple distinction.
On the one hand, the users.
On the other, applications.
The first have an identity, permissions and responsibilities. The seconds carry out processing according to predefined rules.
All modern digital governance has been built around these two categories.
The arrival of AI agents changes the situation.
An agent is not a collaborator. However, he can act on behalf of a collaborator.
An agent is not a classic application. However, it can access several applications simultaneously and make certain decisions within a defined framework.
This intermediate position creates a gray area that current models struggle to address.
Most organizations are therefore faced with a new reality: they must now govern a third category of digital identities.
And this is precisely where the difficulties begin.
The risk is not AI. The risk is the absence of a framework.
When a new technology appears, companies naturally focus their efforts on uses and potential gains.
AI agents are no exception to this rule.
Businesses are experimenting, automating and seeking to save time. This dynamic is healthy. But it can quickly produce the same effects as those observed with Shadow IT or the first cloud services.
How many agents are currently used in the company?
Who created them?
What data do they access?
What actions are they allowed to perform?
Who is responsible for their behavior?
In many organizations, these questions remain unanswered.
The danger is not the existence of the agents. The danger is their proliferation without an appropriate governance model.
Four principles for governing AI agents
CIOs do not need to completely reinvent their practices. On the other hand, they must adapt them to this new category of digital players.
Four principles seem essential to me.
1. Assign a specific identity to each agent
Today, many agents implicitly inherit the rights of their creator or user.
This approach quickly reaches its limits.
Each agent should have a clearly identifiable identity, a designated owner and an explicit scope of action.
The objective is simple: to know at all times who is acting, in whose name and in what context.
We should never have to wonder whether an action was carried out by a collaborator or an agent.
2. Apply the principle of least privilege
Agents must only have the access necessary for their mission.
An agent responsible for supporting a sales team does not need access to HR data. A marketer does not need to intervene in financial systems.
This principle is well known in cybersecurity. It becomes even more important when entities are able to operate autonomously and at scale.
The question should not be:
“What can this agent access?”
But rather:
“What does this agent really need to accomplish his task?”
3. Make every action traceable
Trust is based on visibility.
Every action performed by an agent should be auditable.
What data was consulted?
What decision was made?
What action was performed?
In what context?
This requirement is not just about security. It becomes essential for understanding behavior, correcting errors and demonstrating process compliance.
An agent who acts without traceability quickly becomes an operational risk.
4. Maintain human supervision
Autonomy should never eliminate responsibility.
Not all processes present the same level of risk. Some tasks can be largely automated. Others require systematic human control.
The ability to define these boundaries will become one of the key roles of CIOs in the coming years.
The question is not about choosing between automation and human control. It involves determining where to intelligently place the cursor.
A responsibility that goes beyond the IT department alone
The mistake would also be to consider the governance of AI agents as a purely technical subject.
The issues concern security, but also compliance, human resources, business management and sometimes even corporate governance.
Who can create an agent?
Who validates their permissions?
Who controls its scope of action?
Who bears responsibility when an incident occurs?
These questions require organizational as well as technological answers.
As with the cloud or data, the most mature companies will probably be those which approach the subject in a transversal manner rather than as a simple IT project.
Govern before suffering
The history of digital shows a recurring pattern. Customs always arrive before the rules that govern them.
Smartphones arrived before BYOD policies.
The cloud arrived before cloud governance strategies.
Generative AI arrived before most usage charts.
AI agents are now following the same trajectory.
The good news is that CIOs still have a window of opportunity. The subject remains emerging. The uses are numerous but still under construction.
This is precisely the moment to define the principles that will guide their development.
Because the real time bomb is probably not artificial intelligence itself.
This is the risk of seeing the emergence, in information systems, of thousands of agents capable of acting, accessing data and taking initiatives without the company having clearly defined the rules which govern their existence.
And as is often the case in governance, organizations that act early will have a decisive advantage tomorrow over those that wait until the problem becomes visible.
