Databricks CEO warns full AI automation remains distant despite agent boom

Caption: Ali Ghodsi, CEO of Databricks, said humans will need to be accountable for artificial intelligence's decisions. (Courtesy Databricks)

Databricks CEO Ali Ghodsi has cautioned that complete automation of tasks through AI remains far off, despite growing corporate adoption of AI agents. Speaking at a San Francisco conference, Ghodsi stated that people "underestimate how hard it is to completely automate a task."

Key Points:

• Human oversight will remain critical for years as AI agents proliferate in workplaces
• Databricks (valued at $62 billion) launched a no-code platform for building custom AI agents
• Current AI agents still make mistakes, requiring human "supervisors" to approve decisions

Ghodsi compared the situation to aviation, where autopilot technology still requires trained pilots: "Why do we still want two pilots in there? ... Given that the AIs occasionally get things wrong, in society, we want somebody to be responsible."

The comments come as:

• Companies like Klarna report AI agents doing work equivalent to hundreds of employees
• OpenAI's Sam Altman suggests AI agents are becoming like junior-level coworkers
• Studies show error rates increase with more complex AI agent tasks

Databricks' new platform allows companies to create agents for HR onboarding, policy Q&A, and other functions without coding. However, Ghodsi emphasized that human accountability remains essential, predicting that for years to come, "we all become supervisors" of AI decisions rather than being replaced by them.