|
The next big AI battleground will be centered on software for managing agents, helping businesses handle the growing suite of AI agents they’re using from different providers, as we reported today. On that front, the older enterprise software firms feel they have advantages over OpenAI and Anthropic. That advantage boils down to software firms’ intimate knowledge of customers’ business processes, which software executives think will be tough for the startups to match. That’s especially true for software that companies use to manage the steps involved in common tasks like processing invoices and setting up new employees, known as “workflows.” In interviews this week with ServiceNow and Atlassian, executives highlighted their firms’ deep roots in this area and said this knowledge is crucial for agents to work properly. The focus on agent-management software follows OpenAI unveiling a preview version of its product, called Frontier, last week. In its announcement, OpenAI implied that Frontier could become the primary way companies use software, in essence by directing AI agents through the ChatGPT interface to use a variety of other companies’ apps, such as for HR, IT, or sales tasks. Many established enterprise software companies saw Frontier as signaling OpenAI’s intention to usurp those companies’ relationships with customers, multiple software salespeople tell us. Frontier is limited to a small group of large customers, so it’s not clear how aggressively OpenAI will pitch the product to customers that are using its AI models and other products. But cloud database provider Snowflake, which recently inked a $200 million deal to use OpenAI’s models and other products, and sells a similar product, thinks customers should choose their own management software. "OpenAI is an important partner and innovator, but in the enterprise, we don’t believe any single company should own the agentic control plane,” said Baris Gultekin, VP of AI at Snowflake, said, in an emailed statement. As OpenAI tries to parlay its AI model success into managing agents, the old guard is ramping up its response. In an email to sales teams, Judson Althoff, CEO of Microsoft’s commercial business, asserted that OpenAI has yet to establish itself as a “platform” company that sells many different kinds of products, as we reported last week. Althoff also said Microsoft enjoys other advantages like owning its own infrastructure (Microsoft sells a product called Agent 365 that competes with OpenAI Frontier). Part of the animus around OpenAI’s release is the implication that ChatGPT will become the singular, authoritative dashboard that human employees will use to manage all of their AI agents. That seems to clash with the vision that other firms like Microsoft, Salesforce, and ServiceNow are pitching, where customers use those firms’ dashboards to control agents. Presumably, in the war for dashboard dominance, only one winner will emerge. Still, incumbent firms feel they have an advantage because they already own the applications where customers currently store and organize data related to back-office tasks like HR and IT, while OpenAI doesn’t. “We provide much more than a dashboard. You don’t care where you see the information, but for us it‘s where you manage the full lifecycle of these agents from the automation, budgeting, who gets to build them, how do you deploy them, who’s accountable to making sure they're working,” said Dorit Zilbershot, ServiceNow’s vice president for AI product management. “Thats not something you get from just foundation [AI] models alone.” No one seems willing to say it publicly, but it’s a good bet that sales leaders at other enterprise software firms are saying similar things about the challenges OpenAI and Anthropic could face in getting companies to standardize on their agent management software. Databricks Preparing for a Crash? Enterprise AI software provider Databricks just raised $7 billion of fresh capital—including $2 billion of debt—after reaching a $5.4 billion annual revenue pace last quarter, representing growth of more than 65% from last year’s quarter. The 13-year-old company is now valued at $134 billion. Yet Ali Ghodsi, co-founder and CEO of Databricks, told CNBC earlier this week that some recent conversations with investors have him thinking that the tech industry could be heading toward another dot-com-type crash. Some investors are worried about circular funding deals in the AI market, he said. These include Nvidia’s investments in cloud provider CoreWeave, which CoreWeave has used to buy more Nvidia GPUs to build out its data centers. The ongoing turmoil in the enterprise software market, where some investors seem to think agents are going to replace many of the features that customers have long paid for, or replace entire applications like CRM, is also playing into Ghodsi’s thinking these days, he told CNBC. While usage of Databricks’ products is on the rise as more knowledge workers get access to AI, Ghodsi said investors also seem to be concerned that this could negatively affect traditional software providers and database firms.
|
