Following Palantir's pioneering approach, OpenAI, Anthropic, Microsoft, and Amazon have poured billions into hiring so-called frontline deployment engineers (FDEs) to bring new AI technologies into enterprise client operations. Google, however, is taking a different path: the company says AI can automate a portion of that FDE work on its own.

Andy Gutmans, Vice President and General Manager of Databases at Google Cloud, said that although Google Cloud recently announced plans to hire hundreds of engineers to help clients build applications on the Gemini large model, Google is now leveraging AI agents to automate part of the consulting work these specialists do for large enterprises, primarily focused on organizing and structuring corporate data.

In the past, frontline deployment engineers needed to be stationed on-site with clients, organizing proprietary internal data so that AI could better understand it, enabling employees to complete data analysis faster and more accurately. For example, different companies define financial metrics like "total revenue" in various ways, and FDEs relied on manual processes to ensure AI could recognize and understand these definitional differences.

Gutmans said Google has recognized the limitations of this highly labor-intensive model. "If you want to activate and utilize all of a company's data, you will never be able to hire enough people to do it." In his view, to maximize the value of AI agents, companies should open up as much of their data as possible to the agents, including various documents and legal contracts.

Google's AI agents scan and organize all of a client's data, clarifying the relationships between that data and various business functions such as sales and inventory management. This process generates data with rich contextual information, including knowledge graphs and semantic layers. With this foundation, the workload and cost required for AI agents to handle multi-step tasks, such as invoice processing or new employee onboarding, can be significantly reduced.

This column has previously covered these critical data layers. According to a Google Cloud spokesperson, British telecom and media company Virgin Media O2 used Google's AI agents to integrate 20,000 separate data sets—work that would have taken thousands of hours if done manually. The integrated datasets allow the company's own AI agents to quickly access the relevant business data they need.

Gutmans noted that earlier this year Palantir launched an AI version of the FDE with some similar functions, but Google Cloud views its AI agent for generating business context as a competitive advantage. This agent is part of Google's Knowledge Catalog data management product. The Google Cloud spokesperson added that Google Cloud is developing the AI agent in collaboration with Alphabet subsidiary DeepMind to build "as complete a business context as possible," and unlike many competitors, Google has its own proprietary Gemini large model, which can be used to troubleshoot and verify issues that arise during agent operation.

Gutmans acknowledged that Google's context-generation AI agent is not foolproof, and clients still need to assign a small team to review and validate AI outputs. Google Cloud expects this process to become fully automated in the future. "At this stage, humans still need to step in occasionally to confirm or reject results. But as new data continues to flow in, the system can eventually run entirely on its own through agents."

Note: For learning reference only, not intended as technical or investment advice.