Built In

An interesting thing is happening in policy circles right now: Lawmakers across the political spectrum are starting to ask the same question about artificial intelligence (AI), even when they disagree on almost everything else policy-wise. Who actually benefits from the infrastructure that made this AI boom possible, and does the public have a legitimate claim on some of the value being created?

AI is rapidly changing the way that employers upskill and train employees for the future — and for good reason. A recent Deloitte report found that despite 50 percent of workers having access to AI, only 34 percent of employers are truly using the technology to reimagine their businesses and find success in this new AI-first era. The same survey found that 48 percent of respondents cited AI upskilling as an important part of their talent strategy.

At its annual conference, Google introduced what it described as the biggest update to its search box in more than 25 years, turning search into an AI-driven experience that answers questions, builds visualizations and runs background agents that track information for the user. The familiar list of links is still there, but it now sits within a broader interface that often handles the navigation itself.

When it comes to employee wellbeing at Superhuman, balance isn’t the goal — yes, you heard that right. Sydney Wu, director of people ops and compliance, said that phrase sets people up for failure. “We aim for awareness — knowing when to sprint and when to recover and making both intentional,” Wu said. Built In spoke with Wu about how the AI productivity company ensures its employees stay sharp and engaged while taking the time they need to rest, recharge and return refreshed and ready to work.

You’re about to graduate into a market with jobs that don’t quite exist yet, holding a $200k piece of paper that means less than it should. And at the same time, you’re expected to walk into a company and be the most AI-fluent person in the room. The same institutions that spent your last few years treating AI like a form of cheating are now handing you off to employers who assume you've got it figured out.

SpaceX’s record-breaking IPO has officially made CEO Elon Musk the world’s first trillionaire, reigniting conversations around the growing pay gap between company leaders and the rest of the workforce. According to a compensation survey by the Associated Press, the average CEO compensation package increased 6 percent in 2025 to $17.7 million. Meanwhile, wages have barely kept up with inflation since 2020, leaving many workers feeling like they’re falling behind amid a rocky economy.

An AI agent at Meta was asked to help manage an employee’s inbox. Instead, it deleted the entire thing. At Amazon, an internal agent autonomously decided to tear down and rebuild a deployment environment, knocking an AWS service offline for 13 hours. These are not hypothetical scenarios. They’re reported incidents from this year. What makes these failures instructive is not that the agents malfunctioned in any traditional sense. The agents executed exactly what they inferred they were supposed to do.

Layoff announcements increasingly point to AI as the primary driver. The narrative is familiar: AI will automate tasks, reduce headcount and help companies achieve more with less. But the real cost of AI layoffs may not appear during the layoff itself but through the reentry phase that follows. AI adoption is working, but companies are likely underestimating the future cost of rebuilding the human systems they are quickly dismantling.

Most people think negotiation is about winning. After years on both sides of the table as a venture capital investor, I’ve come to believe the opposite. A good negotiation leaves neither party with exactly what they wanted, yet both walk away trusting the person across the table. That reframes matters because the negotiations that shape a career are rarely one-off transactions.

Scott Shambaugh, a volunteer who helped maintain an open-source library, approved and rejected many requests for the popular Python project, Matplotlib, but this time was different. Scott had noticed a change request had come from an autonomous agent, so he dismissed it quickly.