Companies are increasingly finding ways to automate certain job functions with AI agents, but that doesn't mean that humans will go anywhere soon.
Ali Ghodsi, co-founder and CEO of data analytics and AI startup Databricks, told reporters at a meeting in San Francisco on Wednesday that she believes “people underestimate how difficult it is to fully automate tasks.”
The CEO said humans have been in the loop for a long time since adding levels of oversight and accountability for AI decisions.
“I think in a few years, there will be agents in a lot of places, but there are people who oversee and approve all the steps. If they approve, they will be on the hook and accept 'OK',” he said. “We're all going to be supervisors.”
AI agents are essentially digital assistants deployed by companies to support a variety of job functions, from management tasks and customer service to coding and research.
Klarna said last year that customer service AI agents can do the work of around 700 human employees. In June, Openai CEO Sam Altman said AI agents were beginning to act like junior-level employees.
Wednesday, Databricks, that It has been rated in $62 billionAnnounces a new platform for companies to create customized AI agents without writing Any code.
Ghodsi told reporters that Databricks' customer base is looking for agents who can help onboard HR or answer questions about the company's policy.
The CEO said that using AI agents multiplying in organizations does not mean that humans will be replaced entirely because agents are still making mistakes.
Patronus AI, a startup focusing on LLM evaluation and optimization, has found that the more steps an AI agent takes, the higher the error rate.
Ghodsi pointed to the aviation industry as an example of humans remaining “in the loop” despite autopilot technology.
“Why do we still want two pilots out there? And why do we want pilots to train? We don't want them to train in the cockpit, sleep or get drunk,” he said. “I think it's only a few orders of magnitude hard to get rid of the last mile, and given that AIS sometimes makes things wrong, society wants someone to take responsibility.”
