HashiCorp’s big bet: Infrastructure for the agentic era

HaschiCorp, a software company recently acquired by IBM, hosted its 10th conference in San Francisco this week. From the HashiConf stage, executives made the case that the technology landscape isn’t that different in 2025 from what it was in 2015. Sure, the conversation (and interest) has moved from cloud to AI. But just like in the days of cloud, many enterprises stop short of embracing the full potential of technologies.

“Enterprises want AI to transform their business, but they’re bolting it onto legacy workflows,” said Armon Dadgar, the company’s CTO and Co-Founder, in an interview with IBM Think. “If your process is emailing a thousand people, AI won’t fix that.” Without automation and visibility, AI remains a toy: great for demo, but impossible to use in production.

The path to AI adoption in enterprise can look a lot like the path to cloud adoption. “When we saw people go to cloud, at first, they treated it as ‘Hey, it’s just a managed data center,’” Dadgar said. “‘I’m just going to take my same mental model and the way I operate in my data center, and I’m going to move it to cloud.’ What we saw was that if that was your approach to cloud, you didn’t get any benefits.”

HashiCorp identified an opportunity to deliver a unified control plane that extends across the hybrid cloud. And that’s exactly what it aims to solve with Project infragraph, a tool that allows companies to create a real-time infrastructure graph that connects infrastructure, applications, services, ownership and policy—and in the process, helps lay the groundwork for agentic infrastructure.

Dadgar stressed that as enterprises work to operationalize AI and prepare for agentic workflows, the need for efficiency, security and visibility is more pressing than ever—and by providing a unified system of record for infrastructure and security, Project infragraph can help teams to automate, enforce governance and scale AI safely.

“We’re not trying to solve every layer of AI,” said Dadgar. “But for the infrastructure management layer, the question is: how do we pull all of it together and have a single consistent view of everything? Because if you don’t have that, it becomes very difficult to build these kinds of AI-driven workflows.”

To that end, HaschiCorp also announced lifecycle management updates and lifecycle security enhancements.

For Bruno Aziza, a VP of Software Product Marketing at IBM, Project infragraph could address the gap between experimentation and production. “Many organizations today run on more than one cloud, and there is no effective system to give a first-level knowledge of how all the dots connect with each other,” he said. “That’s why I like the graph idea. But there’s also intelligence on top of that. We get a system where now we know that it’s going to run efficiently, but maybe at some point, you’ll be able to anticipate the issues that might be coming up.”

Above all, Dadgar believes that enterprises still need to ask themselves the basic yet essential questions when thinking about AI.

“The biggest thing is: do you have the foundation in place? I go into a lot of these enterprises where they’re like, ‘I really want to go to cloud or adopt AI.’ But they haven’t built the actual foundation,” he said. “To do AI right, you have to start by doing the cloud layer and the automation layer correctly.”