Two CEOs of multibillion-dollar AI companies say AI won't automate tasks as easily as many leaders assumed.
Glean CEO Arvind Jain and Databricks CEO Ali Ghodsi said on Tuesday's episode of “Bg2 Pod” that companies need to temper their expectations about how quickly and easily they can deploy AI.
Jain said Glean has tried to automate internal workflows, including using AI to automatically identify employees' top priorities for the week and document them for leadership.
“We have all the background within the company to make that happen,” Jain said, adding that AI will “magically” do the job. The idea seemed simple, but it didn't work.
Glean, an AI startup that helps employees search company tools and documents, announced in September that it had raised $150 million at a valuation of $7.2 billion.
Jain pointed to Glean's other failed bet: building and fine-tuning custom models for specific use cases within its products. That effort “didn't really work,” he said, and ended up reverting the company to an existing underlying model that was easier to deploy.
“It actually takes a lot longer than you think to actually create success,” he added.
“Just unleashing an agent doesn't mean it's going to work,” said Godoshi, whose company sells data and AI platforms.
Enabling AI to work within an organization is an “engineering skill” that requires careful evaluation, production work, and a strong team to support it, he added.
Databricks announced last week that it had raised more than $4 billion in a funding round, valuing the company at $134 billion.
Both CEOs said failure of AI projects is common and not necessarily a sign that something is wrong.
“I hear that 95 percent of projects fail,” Jain says. “That's actually what you want.”
“If you're actually experimenting with new technology and all your projects are failing, that means you're not trying hard enough,” he added.
Humans also need to keep up to date with the latest information.
Godi previously said that even as companies deploy more agents and AI automates more tasks, human oversight will continue to be essential for AI systems.
“In a few years, yes, there will be agents in many places, but humans will be overseeing and approving every step. And when you approve, when you click 'OK,' you're going to be at risk,” he said at a conference in San Francisco in June. “We all become supervisors.”
Other technology industry leaders have expressed similar views.
Research scientist Joshua Bengio, one of the “godfathers of AI,” said human qualities will become more important as machines take over jobs.
“It's about striving to be the beautiful person you can be, and even if machines can do most of the work, I think that part of us will still be there,” Bengio said on last week's episode of the podcast “The Diary of a CEO.”
“I think the human touch will become increasingly valuable as other skills become increasingly automated,” he added.
