Agentic AI adoption raises new challenges for enterprise data resilience
The expansion of agentic artificial intelligence is testing companies’ ability to monitor their systems and recover information when an agent acts unexpectedly.
Eighty-eight percent of companies are using or testing artificial intelligence agents, but only 40% of those responsible for these implementations are confident in their ability to accurately isolate and reverse failures involving these systems. This gap reflects the challenges organizations face in monitoring technologies that access critical information and applications and can execute actions autonomously.
Against this backdrop, Veeam Software introduced Agent Commander, a solution designed to address the risks that agentic artificial intelligence poses to enterprise data. The offering aims to facilitate agent identification, monitor their interactions with information, and support responses to behaviors that deviate from expectations.
The difficulties in responding to these incidents add to existing limitations in data recovery. Veeam’s report Data and AI Trust Gap indicates that, although the adoption of AI agents is growing, confidence in the ability to reverse failures involving these systems remains limited.
Meanwhile, the Data Trust and Resilience study reveals that only 28% of companies affected by ransomware managed to fully recover their compromised data.
The challenges of data resilience.
The speed at which AI agents can act on large volumes of information poses challenges that extend beyond traditional data protection and recovery models. An apparently minor deviation can spread rapidly and affect data, systems, and processes that are critical to business operations.
In this context, organizations need to understand how these systems interact with their information, detect anomalous behavior, limit its scope, and respond precisely when an incident occurs.
This requires complementing data protection and recovery with capabilities to monitor the actions performed by agents.
“In the agentic era, organizations need to evolve from a reactive view of data protection toward an approach that allows them to understand what AI is doing, what information it is acting on, and what happens when its behavior deviates from expectations. It is not just about controlling agents, but about having the capabilities needed to detect, protect, and counter risks before an anomaly causes a greater impact,” said Javier Castrillón, Territory Sales Manager at Veeam in Colombia.
Three capabilities for managing risks.
Agent Commander proposes an approach based on three capabilities to manage interactions between AI agents and enterprise data:
Detect: identify which agents exist, what data they use, which systems they access, and when their behavior deviates from established parameters.
Protect: apply controls, policies, and access restrictions to reduce data exposure to individuals, applications, and AI agents.
Undo: respond to incidents and accurately recover clean, reliable information without relying exclusively on explanations provided by the agent or the affected system.
Governance amid adoption.
The expansion of agentic AI also raises questions about organizations’ ability to implement effective controls. Although companies are increasing their investments in artificial intelligence governance, a gap remains between established policies and their operational implementation.
In this context, Gartner forecasts that more than 40% of agentic AI projects will be canceled before the end of 2027 due to rising costs, a lack of clarity about their business value, or insufficient risk controls.
This forecast highlights the importance of developing governance frameworks that go beyond establishing what agents are allowed to do. They must also enable organizations to identify agents, monitor their activity, restrict their access, and respond when their actions deviate from established parameters.
As AI agents move beyond the experimental phase and begin to participate in increasingly important business processes, data trust becomes more important for organizations seeking to expand their use.
The adoption of agentic artificial intelligence thus introduces new requirements for business resilience. In addition to backups and access controls, organizations need visibility into how these systems interact with their information and the ability to respond when their behavior deviates from expectations.
|
|
