|
Welcome back! Last year at Google Cloud's Next conference, executives touted the power of the company’s AI models for businesses. This year the theme is how to help companies use the models. And with good reason: In interviews at the conference, customers and Google Cloud resellers told me that companies racing to adopt AI are hitting roadblocks. Some are still setting up their first AI agents, while others are struggling to manage a menagerie of them. Three years into the generative AI boom, the abilities of tools from Google and other leading AI labs are outpacing the abilities of many companies to understand and implement them, particularly as agents and other AI offerings get more complex. AI companies are having to invest in not only the technology but the tools and specialist consultants and “forward-deployed engineers” to help clients use it. Google on Wednesday announced a $750 million fund for consulting firms to speed up their customers’ AI adoption. “For big companies, the gap between what companies are able to do and what the technology allows is getting bigger and bigger,” said Dave Williams, chief information and digital officer of Merck, the pharmaceutical giant. The biggest roadblocks to broader AI adoption are preparing data and processes, training employees and finding the right tools to use, he said. Merck had already adopted AI in specific areas including scientific research. On Wednesday, it announced a $1 billion deal with Google in which the tech giant’s engineers will work closely with Merck to deploy agentic AI across research & development, manufacturing, commercial and corporate functions. Chris Sakalosky, Google Cloud’s vice president of strategic industries, emphasized that the company is trying to make it easier for customers to adopt generative AI without the extensive support that Merck is getting. One of Google’s main AI announcements at Next was its Gemini Enterprise Agent Platform, a revamped product with new abilities to build and manage automated AI, including new security features and a central registry for an organization’s agents. “The fast adopters figured out some of these now common patterns [of AI use], but they had to crawl over glass these last two years,” said Michael Gerstenhaber, a VP of product at Google Cloud overseeing Google’s AI offerings for businesses. “We need to make those patterns very common for everybody else in order for the technology to fulfill its promise.” Challenges include figuring out what AI agents should have access to and how that access should be distributed across employees, said Karthik Kripapuri, CEO of Promevo, one of the Google Cloud resellers that help companies adopt AI products. And many companies implementing AI are starting to see “agent sprawl,” as employees create and adopt different agents, said Beau Broker, field CTO for the western U.S. at AHEAD, an IT consulting company. Like other companies, the Home Depot has been struggling to adapt full workflows to AI, but it has found success with specific AI solutions for tasks such as coding or customer engagement, executive vice president Jordan Broggi said. The home improvement retailer announced new AI-powered customer support phone agents, powered by Gemini Enterprise, on Wednesday at the conference. It already has given all corporate employees access to Gemini and Microsoft Copilot, and its software developers have flocked to AI tools from Google, Anthropic and OpenAI—switching between models as new releases leapfrog competitors in quality, Broggi said. He likened the situation to a factory floor: Using a robot at one station in a factory only speeds up the assembly line until the next step. The Home Depot has automated some of the steps, but not yet all of them, he said. “You just move the constraint to the next place, and so you really have to get better and faster across all the steps,” he said.
| 
