Most companies adding AI to their workflows want to know how fast it can work, how much more  efficient it will make their workforce, and ultimately how it will affect their bottom line. The better question starts with whether those old workflows make sense in the first place.

That distinction matters now more than ever. Challenger, Gray & Christmas reports that AI was the top-cited reason for U.S. job cuts in March 2026, making up  25% of all announced layoffs. That number was 5% in 2025. Companies are cutting because the machine can do the output. The work still gets done. The cost is real.

The result is a volatile environment for employees, a brutal job market for the people cut, and growing public distrust of AI itself. I think the companies making those cuts are getting the AI question wrong in the first place. "AI can handle it" is not a strategy. It is a shortcut. And shortcuts are what you take when you have not actually thought about where your business needs to go in the next ten years.

Here's the question I think every founder and operator needs to sit with: If a model ten thousand times more prolific than your best employee joined your company tomorrow, what would you have your team do instead?

Leadership should look at this problem and instead of cutting headcount, shift those responsibilities and point people at the problems that don’t have a solution yet. Let people use their superpowers, thinking differently, while AI handles execution on what’s already understood. 

New research from MIT Sloan shows the real gains from AI don't come from accelerating individual tasks,  they come from restructuring how tasks are sequenced and handed off. Speed up one step in a ten-step process, and you've moved the bottleneck and freed up time for other tasks. The companies getting the most out of AI are building new processes and new capabilities with the time earned back from AI, not just making more of the same thing.

In the right workflow, AI compresses work that once required weeks into hours. I've seen it. Napster builds products that depend on it.

What it can't do reliably is produce a genuinely new idea. That’s the human element that can’t be recreated.

AI is very good at pulling from what exists, identifying patterns, and leveling up outputs based on previous iterations of those outputs. That's useful. But it's not the same as the kind of thinking that actually moves a business forward. Wharton researchers Gideon Nave and Christian Terwiesch studied this directly and found that while AI improves the quality of individual ideas, it collapses creativity over time. 

In this study,  researchers gave AI and human participants a task: invent a toy using a brick and a fan. All of the human-made toys were unique. Only 6% of AI-generated ideas were; the rest were “clustered around the same concept,” in some cases even down to the name of the toy. 

That’s what happens when everyone uses the same tools with the same foundations the same way. Sameness, even an unprecedented amount of it, does not move companies forward for long.

There's a reliability problem underneath the creativity problem, too. A Princeton study found that despite rapid gains in AI accuracy on benchmarks, reliability and consistency have barely moved. AI still struggles to predictably answer the same question over time, even when the question is asked the exact same way. It can be accurate, but it cannot yet be dependable in the way a human employee is. 

Organizations making workforce decisions based on benchmark scores are going to find that out the hard way, if they haven’t already.

So what do people do when an AI outpaces them? They focus on the domains only humans can execute reliably: creativity, judgment, and dependability. The things AI can’t reliably pattern-match. The things that make us uniquely human.

The companies cutting headcount to capture AI efficiency are making a short-term bet that execution is the business. It isn't. Execution is now a commodity, and that just makes what humans bring even more valuable than before. Because sameness is also commoditized, and building differently is the only way to stand out when everyone has the same tools at-hand.

The models will get better. They might even stretch toward some of these more human-like functions. That part isn't in question.

But what stays irreplaceable is the person who can think beyond its foundations. We have plenty of those people at Napster, and we’re hiring more to meet the AI moment. 

How about you?