AI CEOs say automation with AI is harder than it looks

Two prominent CEOs at the forefront of the AI industry have issued a stark reminder about the complexities of automating workplace tasks. Arvind Jain, CEO of Glean, and Ali Ghodsi, CEO of Databricks, shared their insights during a recent episode of the "Bg2 Pod," highlighting that the deployment of AI is not as straightforward as many corporate leaders might believe. Jain recounted his attempts to streamline internal processes at Glean, which specializes in helping employees navigate internal tools and documents. He aimed to utilize AI to automatically identify and document weekly priorities for staff. Despite the company's extensive context and resources, Jain admitted the implementation did not yield the expected results. "I thought AI would just magically handle it," he said, but the reality proved to be more complicated. Glean recently secured $150 million in funding, boosting its valuation to $7.2 billion. However, Jain acknowledged that another initiative—developing a custom AI model tailored to Glean's specific needs—also fell short. This experience led the company to revert to existing foundation models, which are simpler to deploy. "Success takes much longer than anticipated," he remarked. Ghodsi echoed Jain's sentiments, stating that effectively integrating AI into business processes is not as simple as letting "agents" run autonomously. He described the process as an "engineering art," requiring thorough assessment, consistent production efforts, and robust support teams. Databricks recently raised over $4 billion in funding, valuing the company at an impressive $134 billion. Both leaders acknowledged that setbacks in AI projects are commonplace and should not necessarily be viewed as failures. Jain noted that a high failure rate—often cited as 95%—is actually indicative of an organization pushing boundaries and experimenting with new technology. "If all your projects succeed, then you might not be trying hard enough," he suggested. Ghodsi emphasized the ongoing necessity for human oversight in AI operations, even as automation becomes more prevalent. He predicted that while AI agents will proliferate in various sectors, human intervention will remain crucial at every step. "We will all take on supervisory roles," Ghodsi stated, reinforcing the idea that accountability will still lie with human operators. This perspective is shared by other tech visionaries, including AI pioneer Yoshua Bengio, who believes that the value of human qualities will increase as machines take over more tasks. "Focus on becoming the best version of yourself. Even as automation rises, the human touch will gain greater importance," Bengio remarked, reflecting on the evolving relationship between humans and technology.