At VentureBeat's AI Impact event, Brian Gracely, senior director of portfolio strategy at Red Hat, detailed what separates enterprises that scale agentic AI from those stuck in pilot purgatory. He focused on cost discipline, security blind spots unique to autonomous systems, and the organizational friction that determines whether agent adoption spreads beyond early champions.
Enterprises overestimate how far behind they are on AI agents
Many executives, following industry keynotes and AI announcements, fear they are already dangerously behind competitors deploying agents at scale. Gracely argues this anxiety is largely misplaced: teams move up the learning curve much faster than expected once they start building. However, rapid progress creates a different challenge: as agent usage expands, AI costs rise just as quickly, turning cost management from an engineering concern into a recurring boardroom discussion.
Sponsored Protocol
Right-sizing AI models is the fastest lever for cutting agent costs
The biggest cost issue is that enterprises overspend by defaulting to the most capable model regardless of task complexity. "If I'm simply trying to resolve an insurance claim, I don't need to know about the history of Western civilization or World Cup soccer scores," Gracely said. To address this, many companies adopt semantic routing, which automatically classifies requests and sends each to a model sized for the task. Infrastructure techniques like caching repetitive queries further reduce the need for GPU compute. According to Gracely, these tools remove the assumption that efficiency and innovation are binary choices.
The financial discipline required for token spend mirrors the FinOps practices that took years to mature in cloud computing. The underlying frameworks will transfer even as the vocabulary changes, Gracely noted, especially as organizations push for internal education on model selection so teams stop defaulting to the most expensive option. "The same way we first had to teach financial people what an EC2 instance and an S3 bucket are, you're going to have to start explaining tokens to them," he said. "We don't always need a Rolls-Royce. We don't always need caviar, because we're trying to do basic things."
Sponsored Protocol
Patch speed becomes critical as AI tools find vulnerabilities faster
AI-powered vulnerability discovery is forcing enterprises to rethink how quickly they identify, validate, and deploy patches. Traditional patch management cycles may no longer suffice when AI can uncover – and attackers exploit – new vulnerabilities much faster. "Most companies will probably have a window of 7 to 14 days to stay ahead," Gracely warned. "Groups, including Red Hat, will build patches, but the embargo window will be short." AI is also changing what defenders must look for: not just isolated critical flaws, but combinations of minor vulnerabilities that become dangerous when chained together. As software complexity and vulnerability discovery accelerate, the ability to rapidly manage and update software becomes a strategic capability.
Sponsored Protocol
For a real-world example of how dependency on few providers creates risk, see the analysis on China's pursuit of Nvidia chips despite US sanctions, highlighting market concentration dangers. On the security front, the recent Nextcloud leak exposing 367,000 records underscores the need for swift vulnerability management.
Subject matter experts and compliance teams decide whether agents scale
Ultimately, organizational adoption hinges on deep, sustained involvement from subject matter experts whose knowledge the agent encodes. Earning their buy-in is a prerequisite, not an afterthought. "You have to think about incentives, what you do for people who participate so they don't feel threatened that the agent will take their job, and how you incentivize long-term cooperation with innovation," Gracely said. Compliance teams also must validate that agents operate within regulatory boundaries. In the end, success for enterprise AI agents depends not only on technology but on organizational culture and financial discipline, as outlined in Wikipedia's entry on FinOps, a cost management model increasingly relevant for AI.
Sponsored Protocol
