Real Challenges of Enterprise AI Agents: Cost, Security, and Culture

Enterprise AI agents are transitioning from pilot projects to industrial use but face significant challenges. As Brian Gracely, Director of Portfolio Strategy at Red Hat, noted, companies often overestimate their lag behind competitors while underestimating the complexities of scaling. The rapid growth of AI costs has become a key issue, turning expense management into a board-level concern.

The primary cause of rising costs is inefficient model selection. Enterprises frequently use powerful general-purpose models for simple tasks, such as processing insurance claims, where specialized solutions would suffice. Semantic routing and caching of repetitive queries help reduce GPU load and optimize expenses. Experts compare this situation to the evolution of FinOps for cloud services, where financial discipline became an integral part of IT strategy.

AI is also transforming cybersecurity approaches. AI-powered tools detect vulnerabilities faster, reducing the time for exploitation by malicious actors to 7–14 days. This requires companies to accelerate patching and revise security management strategies. Additionally, AI identifies combinations of minor vulnerabilities that become dangerous only when used together.

Organizational barriers remain a critical success factor. Scaling AI agents requires active participation from domain experts whose knowledge is encoded into systems. Companies must create incentives for employees to overcome resistance to innovation and job displacement fears. Without this, even technically advanced solutions risk remaining underutilized.