A new Google Cloud report finds that 43 percent of IT leaders see trouble linking legacy APIs and data sources as the biggest barrier to scaling agentic AI. The State of Infrastructure in the Agentic AI Era report shows enterprises struggle more with supporting systems than with the models themselves. Eighty-three percent of organisations say they need infrastructure upgrades for production-grade agentic AI. Eighty-one percent point to operational complexity and engineering overhead as major unexpected costs. Thirty-six percent say a lack of specialised high-throughput vector databases limits the business context they can feed to AI models. Agentic systems need continuous access to data across databases, analytics tools and older applications. Without enough context, agents produce incomplete or inaccurate results. Google Cloud argues firms should avoid simply moving large datasets into AI platforms. That approach raises complexity and cost.
A new Google Cloud report finds that 43 percent of IT leaders see trouble linking legacy APIs and data sources as the biggest barrier to scaling agentic AI. The State of Infrastructure in the Agentic AI Era report shows enterprises struggle more with supporting systems than with the models themselves. Eighty-three percent of organisations say they need infrastructure upgrades for production-grade agentic AI. Eighty-one percent point to operational complexity and engineering overhead as major unexpected costs. Thirty-six percent say a lack of specialised high-throughput vector databases limits the business context they can feed to AI models. Agentic systems need continuous access to data across databases, analytics tools and older applications. Without enough context, agents produce incomplete or inaccurate results. Google Cloud argues firms should avoid simply moving large datasets into AI platforms. That approach raises complexity and cost.