Uber is Hiring More Engineers Because AI is Making Them More Valuable, CEO Says

Uber is hiring more engineers rather than fewer because AI tools have made them "superhumans," CEO Dara Khosrowshahi said, pushing back against the industry trend of using productivity gains to justify headcount cuts. Speaking on the "On with Kara Swisher" podcast, Khosrowshahi noted that other tech executives see AI making engineers 20% to 30% more productive and conclude they need 20% to 30% fewer engineers. His view: every engineer has become more valuable. Between 80% and 90% of Uber's developers now use AI tools, according to Khosrowshahi.

The company no longer keeps scores of engineers on call to diagnose issues because AI agents are constantly monitoring systems, he said. The latest AI models are producing "hundreds of millions of dollars of benefit" for Uber, he said, describing the company as an "applied AI" business that harnesses the technology for pricing, payments, matching, routing, identification and customer complaints.

Hiring more engineers because of AI

[Comboman]

Likely they are hiring more engineers to the fix the "vibe coding" mistakes made by managers with AI tools.

Re:

[ghinckley68]

yea no shit cursor ai the code it make is just shit. I use it because i can do an entoire class and all the crud and screens to go with it in a day or two max and the code can be a lot mos sfisticated. I just clean up the mess and fix. but good god some one who does not have 20 years of exp to know what is good and what is crap.

Re:

[gweihir]

As some people are making very good profits with fixing vibe-"coding" crap now, that is entirely plausible.

Makes sense

[GameboyRMH]

If AI makes a programmer appear more productive instead of slower, [andrewzuo.com] it's means they're slapping code together with great haste producing a ton of bugs, and they're going to need more engineers to fix those bugs, thus increasing headcount and showing how AI improves productivity and saves mon- HEY WAIT A MINUTE

Re:

[alvinrod]

AI can spit out a lot of boilerplate type code (some of it close enough to correct) that it can replace the sort of code monkeys that were slapping together bits of code sourced from across the internet all day and hoping it would work before seeing if adding another snippet from elsewhere might solve their problem. The LLMs just do this more efficiently.

A skilled developer could use an AI to generate most of this and fix the mistakes well enough that they're more productive than the aforementioned doork

Re:

[tigersha]

You just described my job right there

Re:

[gweihir]

AI also adds a ton of security vulnerabilities, cannot find most of them and makes attackers a lot more productive. The fireworks will be spectacular.

don't believe it

[dfghjk]

They'll say anything to improve public perception. Also, it's Kara Swisher, the very same grifter that promoted Elon Musk for years. No need for integrity as long as the money is flowing in.

Re: don't believe it

[home-electro.com]

Yeah it is an interesting twist. I guess that is happening they feel pressure to make an impression of adopting AI, and so whatever you do, hire, fire, take a shit, it has to be because of AI adoption.

Expectations of improved productivity is what drives P/E higher, so...

My perspective

[MpVpRb]

I've been writing code since 1972
I use AI as a super reference to get API and function information and produce sample code. This is faster than finding it the old way.
I would never rely on AI generated sample code in my projects. I check every line and ask AI to explain the lines I don't understand.
The final result is written by me, with AI help.
AI can be a great tool to help experts, but if it's used by the clueless, the result is big-ridden slop

Re:

[gweihir]

AI can be a great tool to help experts, but if it's used by the clueless, the result is bug-ridden slop

The problem is most coders were not actually experts before AI. They will be even worse now.

Reasonable explanation: growth over savings

[Somervillain]

This is the traditional economic model. If your engineers are 20-30% more efficient (which I don't believe, but that's irrelevant), nearly every company expands the scope of their ambition and missions. No publicly traded company wants the same volume, but 20% cheaper. They'd rather steal marketshare from their competitors or launch new offerings with this new productivity. Any CEO that blames laying off engineers on AI is lying and should be prosecuted for frauding investors.

If AI could actually cod

yes, in a resource constrained environment

[david1k]

This makes sense for certain types of engineers. However, if you are are a crank-turning-code-monkey... it is a different story. However, with AI, a mediocre engineer can be more reliable and achieve more--freeing up superstars for the hard problems and providing positive feedback that those medium folks would have not gotten. Glad they are making it work. now, stop running stop signs and trying to mow down children at bus-stops.

Re: yes, in a resource constrained environment

[silveride]

I feel of all the companies these guys got it right( not that happy about this company again ). This is a repeat from the 1990s and 2000s. Uber is ahead of the curve. It would be a matter of months before other companies realised what actually is the potential is an ai powered software engineer.

Re: yes, in a resource constrained environment

[home-electro.com]

Do you even code, bro?

You need to be CEO-level stupid to think that AI powered mediocre engineer can be more reliable.