A recent PwC survey found that 79% of executives say their companies already use artificial intelligence agents. However, when looking only at systems that are actually operational, the numbers drop dramatically. A survey by BvLogic found that only 31% of companies have at least one AI agent in production, which explains why many organizations test the technology without fully integrating it into their operations.
From testing to operation: a complex leap
Creating an agent that can perform a task in a controlled environment can be relatively simple. The situation becomes more complicated when the system needs to access internal data, follow corporate rules, and continue to function in the face of unforeseen situations. Good performance in a demonstration does not guarantee that the agent is ready to operate unsupervised in processes involving customers, payments, documents, or internal operations.
Governance and observability are barriers
The more autonomy an agent is given, the greater the need to define clear boundaries. Who is authorized to approve an action? What information can the system access? When is human review mandatory? And what should be done when the tool makes a mistake? According to Kore.ai, 62% of companies have already postponed deployments due to governance concerns, while 53% admitted to putting an agent into production without full confidence in its behavior.
Inadequate evaluation stalls projects
Measuring whether an agent actually performs the intended task goes beyond analyzing a few responses. Limited testing can hide flaws that only become apparent when the system deals with hundreds or thousands of different scenarios. BvLogic notes that the lack of evaluation is one of the main reasons why 88% of pilot projects never reach the production phase. Before expanding usage, companies need to establish objective criteria: which tasks will be automated, what results are expected, which errors are tolerable, and at what point human intervention is essential.
The role of professionals also needs to evolve
The spread of AI agents changes the type of knowledge required in organizations. Professionals involved need to understand processes, identify where automation adds value, monitor results, and recognize situations that require human supervision. This applies to areas such as technology, marketing, customer service, human resources, finance, and operations. As an agent takes over steps in a process, the importance of those who can supervise its work and align it with the business’s real goals increases.
