Applied AI

Following the playbook Palantir pioneered, OpenAI, Anthropic, Microsoft and Amazon are investing billions of dollars to hire so-called forward deployed engineers to get newfangled AI to work in enterprises. 

Then there’s Google, which says AI can automate some of what these FDEs do. 

While Google Cloud recently said it plans to hire hundreds of FDEs to help customers figure out what kind of apps they want to develop with its Gemini AI tools, the company is now automating some of the work these specialized consultants do for large companies to organize their data, said Andi Gutmans, a Google Cloud vice president and general manager who oversees its database products.

Until recently, FDEs worked on-site with customers to organize proprietary corporate data to ensure AI could make better sense of it and thus help employees analyze it faster and more accurately. For instance, businesses often define financial metrics such as gross revenue differently, and the FDE’s manual work aimed to ensure the AI would understand such differences.

Gutmans said that Google has realized the limits of this labor-intensive approach. “If you want to move to activating 100% of your enterprise data, you're not going to be able to hire enough people to make that happen,” he said. 

In his view, customers can get the most value from AI agents by giving them access to as much data as possible, including their files and legal contracts. Google’s AI-powered FDEs scour through customers’ data to understand how it relates to various parts of their business operations, like sales and inventory management. 

This process creates context-rich data, including knowledge graphs and semantic layers, that reduce the effort (and cost) an AI agent must make to handle multistep tasks such as routing invoices and processing new employees. We wrote about those key data layers here.

Virgin Media O2, a British telecommunications and media firm, used Google‘s agents to connect 20,000 separate sets of data, a job that would have required thousands of hours of manual work, according to a Google Cloud spokesperson. The combined data set makes it easier for Virgin Media O2’s own agents to find data that is relevant to their scope of work. 

While Palantir announced a similar, AI-powered FDE earlier this year, Google Cloud sees its own version, Knowledge Catalog, as a competitive advantage, Gutmans said. 

One reason is because Google Cloud works with Alphabet’s Google DeepMind unit to develop “the best possible context” for agents, the Google Cloud spokesperson said. And unlike many of its competitors, Google Cloud has its own AI model in Gemini that it can use to evaluate problems with its agents, the spokesperson added. 

Google Cloud’s context-creating AI agents aren’t foolproof, Gutmans said, so customers must still assign a minimal number of human staff to vet the AI’s work. 

Over time, Google Cloud expects the process to be entirely automated. “There's a period of where humans will occasionally step in to say yes or no. But over time, it can become fully agentic as new data comes in,” Gutmans said.