AI agents are gaining traction in businesses and require new skills
2–3 minutes
Asking a system to consult a spreadsheet, cross-reference information with a customer’s email, and send a summary to a team—all without human intervention at each step—is the kind of task that defines agentic AI. This concept has gained momentum in investment plans, consulting forecasts, and discussions about the future of work.
How agentic AI differs from other tools
Generative AI tools respond to direct commands: a user requests a text or analysis, and the system delivers the result. An autonomous agent, on the other hand, receives a goal, breaks the work into steps, uses resources like calendars, email, spreadsheets, and internal systems, and executes actions with a degree of independence.
In practice, instead of asking for a draft, a professional sets a goal—for example, reconciling monthly releases and identifying discrepancies. The agent processes the data and returns the result for review, which remains a human task.
The adoption of AI agents in businesses
Gartner predicts that by the end of 2026, 40% of enterprise applications will have AI agents for specific tasks, up from less than 5% the previous year. While this projection reflects a growth trend, it does not yet represent current reality.
In Brazil, the IBM Institute for Business Value’s study “5 Trends for 2026” found that 75% of executives expect AI agents to operate independently by the end of this year. Another 65% say these systems already help make decisions more quickly and accurately.
How to prepare to work with AI agents
As agents take on operational tasks, those who can define clear goals, review results, and identify errors become more important. This applies even to those who don’t program, as delegating tasks to a system requires understanding its capabilities.
For those looking to delve deeper into the topic, Saint Paul’s Pre-MBA in AI offers four online practical classes.
Risks and limitations that require attention
Autonomy comes with risks. In IEEE’s global survey on the impact of technology in 2026, excessive reliance and potential inaccuracies (54%) and exposure of internal data (52%) were among the top concerns about using generative and agentic AI in organizations.
That’s why designing an agent involves defining what data it can access and when a person needs to approve an action, such as in payments or customer communications. In customer service, for example, an agent can consult a purchase history and propose a solution, while higher-value refunds require human validation.
What to watch in the coming months
The focus now is on moving from testing to daily operations: which processes will have agents in action, how results will be measured, and what skills companies will start demanding. For professionals, the journey begins with understanding the concept, experimenting with tools in their own work, and knowing when human review is essential.
Additionally, it’s important to monitor how companies will address responsibility for decisions made with the help of agents and what internal rules will be created to record and audit these systems’ actions.
Artificial intelligence is already shaping career plans across various fields
Secure your spot in Saint Paul’s Pre-MBA in AI for R$ 37.
