Why Enterprise AI Success Depends on Infrastructure Not Just Models

AI agents have been part of the conversation for years. But actual use was rare and mostly tied to software development. That changed recently. The rise of personal agents and easier ways to build agent workflows made it feel like the agent age had arrived.

Brian Gracely, Senior Director at Red Hat, shared insights from companies running agents in production. Many leaders worry they fall behind deploying agents at scale. Yet, organizations tend to learn faster than they expect once they start building. Still, as agent use grows, AI costs rise sharply. This makes cost management a hot topic in boardrooms.

Gracely pointed out that the top AI providers are losing money. “The two or three top providers are already telling the market that they’re losing money, and they’re trying to go public to make up those gaps,” he said. Enterprises often overspend by defaulting to the most capable model, even when simpler tasks don’t need it. Gracely said, “We don’t always need a Rolls-Royce. We don’t always need caviar, because we’re trying to do basic types of things.”

To control costs, companies use smart routing. This means classifying requests and sending each to a model sized for the task. Caching also helps by reducing how often requests hit expensive GPU compute. Managing token spend has become like managing cloud computing costs. Teams must learn when to pick smaller models to avoid waste.

AI Adoption Requires More Than Technology

Successful AI adoption isn’t just about tech. It needs deep involvement from subject matter experts and compliance teams. Incentives and long-term cooperation are key. Sloan Session, CFO at Dura Software, said, “The agents handle the pull. The humans handle the judgment and the personal touch.”

Security also drives adoption challenges. AI-powered tools uncover vulnerabilities faster than before. Most companies have only seven to 14 days to fix issues and stay ahead. These tools find dangerous chains of minor vulnerabilities that otherwise go unnoticed. Berry Carter, CEO of S&B Filters, emphasized governance: “If a user cannot access specific information within NetSuite, that user should not gain access to the same information through an AI assistant.”

Shifting Value: From Models to Infrastructure

For the past two years, the AI industry focused on building the most capable model. Now the frontier is shifting. The next big value will come from infrastructure that links intelligence to execution. Enterprise search is evolving into enterprise intelligence. This means synthesizing knowledge and reasoning over proprietary data.

AI is moving beyond standalone assistants. It is becoming a foundational layer within enterprise software. Organizations now invest less in isolated AI demos. They focus more on platforms that integrate knowledge, governance, security, and workflow automation. The competitive edge is no longer just model performance. It’s about reliably deploying and operationalizing intelligence at scale.

Investors watch this shift closely. They evaluate AI companies on their ability to build infrastructure for deployment and governance. The defining question in AI’s future may be who builds the systems that turn intelligence into reliable execution. Many businesses test new AI capabilities through existing processes and needs. Different departments have different AI requirements, causing varied adoption patterns.

Some workers benefit from AI quietly reducing effort. For example, AI shortens reporting cycles. Analysts and operational teams gain from conversational interfaces that help explore data. Employees spend much time hunting for information across fragmented systems. AI-connected workflows automate many manual tasks like revenue reporting and tracking customer backorders.

Most organizations don’t want to remove judgment from workflows. They want to reduce time spent gathering and organizing information. Access to information must remain governed. Permissions, approvals, and security policies stay important. Companies face new questions: which tools to use, who should access them, and how governance must evolve as AI use grows.

Some customers want AI embedded inside operational workflows. Others want to connect their data to external models. Many want both. Oracle NetSuite offers an AI Connector Service and supports Model Context Protocol (MCP). These help connect business information securely to workflows and systems.

The future of enterprise AI is clear. The smartest model won’t win alone. Companies that build strong, secure infrastructure to deploy and govern AI will lead the way.

Stay connected via Google News

Follow us for the latest AI News