Uber CEO on the most promising way to succeed with AI: Throw out the old policies

• Uber CEO Dara Khosrowshahi says some companies are "play-acting" when it comes to AI adoption.
• Speaking at Davos, Khosrowshahi said it's "much harder than it sounds" for companies to embrace AI.
• Companies have been ramping up their spending on AI tools, but with mixed results.

How do you separate the pretenders from the real deal when it comes to business AI adoption?

For Uber CEO Dara Khosrowshahi, it comes down to building AI into processes from the ground up.

Speaking at the World Economic Forum in Davos, Dara Khosrowshahi said some companies are "saying the right words" and "play-acting their way into a pretend transformation."

For example, having an AI agent complete tasks like summarizing a client pitch is the "easy stuff" that won't differentiate companies, he said.

"Truly changing how you work with AI, we've found it to be much harder than it sounds," he added.

He gave the example of using AI to improve customer service at Uber. The company had some success by having AI follow its old policies, but had a real "breakthrough" after its developers rebuilt everything from scratch.

That involved giving an AI agent clear underlying goals, Khosrowshahi said, such as making the customer feel good after the interaction.

"Allowing the AI actually to reason through that and throwing away all of the old policies is turning out to be the most promising way forward," he said.

Khosrowshahi added that companies are essentially a "bunch of policies," and that to get the full potential of AI, they would need to "break down those rules and start over."

Internally, Uber developers are using AI tools like Anysphere's Cursor and Anthropic's Claude, he added.

Many companies have been ramping up their AI spending in the hope of unlocking productivity gains. A recent poll of IT professionals by RBC Capital found that 90% of respondents plan to spend more on AI this year.

However, there are also fears that AI isn't always living up to the hype, and companies are grappling with concerns that it's eroding workers' skills.

Khosrowshahi acknowledged that getting AI adoption right at a company isn't always plain sailing.

"You have to survive through a bunch of car crashes internally to do so," he said.