Tasks that once took six hours now take less than one. A two-week process can sometimes be finished in an afternoon.
But workers aren’t getting their time back.
Instead, executives say companies are using those productivity gains to demand more output from the same employees—turning what used to be an eight-hour workload into something far larger.
You used to spend six hours on that. Now it takes 40 minutes. But nobody is sending you home early. The anxiety gripping corporate America about artificial intelligence (AI) isn’t what you think. It’s not about the machines taking over. It’s about what happens to employees after AI turns their eight-hour workday into two—and the boss still expects them at their desk until closing time.
Many corporate leaders are hesitant to trumpet these wins. “Organizations are a little bit, nervous, is maybe the word,” Ahmad told Fortune. In private conversations with Google, she said, executives admit they are thinking hard about the implications of what all these efficiencies are suggesting.
The nervousness reflects a paradox about a giant leap forward in time savings that turns out to be very real. The question of what replaces that time is not.
Economists and philosophers have been here before. John Maynard Keynes famously predicted in the 1930s that by 2030, a 15-hour work week would be possible—and then asked, with obvious unease, what people would do with all that free time.
Harford traced this paradox to the history of supposedly liberating technologies: email was faster than a letter, but spawned a “profusion of low-quality, low-value messages bleeding into the evenings and weekends.” PowerPoint meant that “highly paid and skilled professionals started wasting time making their own slides badly.”
In other words, the question isn’t whether AI gives you back six hours. It’s whether anyone lets you keep them.
Instead of sending workers home early, Manos said his teams are simply getting more done. A product development cycle tracking to take 24 to 36 months was completed in six months once his team incorporated AI capabilities. Rather than reduce staff, he redeployed those developers to additional projects. “It’s not so much about people are going to lose jobs, or you’re going to sort of shrink that workforce,” he said. “It’s about becoming more efficient and, in our case, getting to market faster.” More capabilities, services, and features will have to be delivered within the same historical timeframe.
Walsh of KPMG agreed with Cappelli’s takeaway, saying that “embedding AI into a business takes time.” Organizations have to “rework all of the process flows,” which includes cleaning up their internal data, aligning all their data flows in the same direction, and doing so across the entire business, across the back office, front office, or middle office, whichever the company is focusing on. Large companies—and those with the capital to invest—have been doing this for the past two years already, he said, characterizing it as just a start. “There’s so much work to be done around this.”
The catalyst for the productivity shift—where it is actually happening—is the evolution of what Google calls the “agentic data cloud,” in which AI models no longer just answer questions but also act as planners and executors. Google’s Gemini 3, for instance, has moved beyond simple Q&A to what Ahmad calls a “thinking role.”
She claimed that the model can first build a plan, then explore multiple approaches, evaluate them against each other, and hone in on the best answer for the customer.
Salva, who spent a decade at Microsoft before joining Google, acknowledged that “we all know that we’re traveling west”—meaning the entire industry shares the same vision of AI autonomy, even if the paths differ. “We’re all trying to get to the same degree of automation. We have slightly different flavors of implementation and workflows for it.”
“The difference we’re seeing, even from six months ago, is organizations are stepping away from small pilots and experiments with generative AI, where they were finding 5, 10, 15, 20-second savings,” he said, “and seeing where an agentic agent is able to actually automate large portions of work entirely so that they can actually reimagine kind of how work is done.”
The back office of an insurance company, he argued, is a prime example: binding a new policy or processing a small business loan currently requires multiple customer interactions, a front-line rep capturing information, a back-office team making a decision, and then a rep communicating that decision back. “These processes generally require either very long conversations or multiple interactions,” Buesing said, offering the examples of a front-facing representative capturing information while a back-office team works on the decision. “AI can perform those functions faster, run a customer history profile in real time while the customer is still speaking to the front-line rep, and help that human make a decision.”
A McKinsey survey of 440 customer experience and operations executives found that 60% or more of the tasks performed in customer operations today are “potentially addressable with AI.” But Buesing was careful to separate the addressable from the capturable. “What is addressable versus what will be capturable, and with what time period? Humans don’t necessarily adapt to change as quickly as the technology is evolving,” he told Fortune.
The new AI voice agents, which six months ago still sounded “tremendously robotic,” have recently crossed a threshold. Latency is barely perceptible, and the agent “sounds casual, fun, friendly, even a little bit joking around.” Early evidence also suggests that, in certain contexts, such as first-round job interviews or ordering sensitive medication, customers actively prefer talking to AI because they “don’t feel judged.”
Buesing said he had independently read the same Harvard Business Review article on work intensity and largely agreed with its premise. Once building AI agents moves from “nights and weekends fun project work” to the expected baseline output an employer demands, workers will “find themselves on a wagon wheel of having to build more agents to try to keep up with the expectations of production,” he told Fortune.
The answer, Richardson suggested, is not to hide productivity gains but to invest visibly in people so they feel equipped for the new regime. “Investing in upskilling is not just a strategy,” she said. “It’s a reassurance. It’s a trust pact between the employer and the worker.” She said companies have a lot of work to do, adjusting to the new mentality of what it means to do work in the AI age. “We need to help reframe productivity for our workers,” she said, because little task completion moments will be swallowed up by AI efficiencies. “To me, it’s shifting from productivity based on volume of work to value [of work], and that’s a big shift within an organization.”
For Salva at Google, who has spent 25 years watching developer tools evolve, the better analogy for where we are isn’t email or PowerPoint. It’s the five stages of autonomous driving, and we’ve only reached stage three or four. The real promise, he told Fortune, isn’t that AI does your job faster; it’s that it changes which parts of the job are yours to do. He said the best developers he sees today aren’t hammering at keyboards—they’re “locked into the architecture,” delegating execution to “a fleet of agents” running in parallel while they hold the big picture in their heads. “That,” he said, “is where productivity happens. That’s where focus and flow happen.”
Where Salva diverges from some of his competitors is in what the future should feel like. “If we’re optimizing for short attention spans,” he said, “what we’re really sacrificing is that delightful Zen moment that you get when you’re locked in”—the deep focus that he believes is where the most important work actually gets done. He said he spends significant time thinking about how to design tools that preserve that state even as they delegate the mechanical work to external systems.
What Manos at Dun & Bradstreet found is that the real disruption isn’t technical, it’s cultural. “At the end of the day, the AI revolution will be successful when you’ve actually changed the people and the people culture to adopt this new framework,” he said. He thinks his company is succeeding where others have failed in AI adoption because it approached things differently. It rolled out AI gradually, starting with small wins: automating the repetitive tasks, like quality assurance testing.
“We didn’t jump in and go, ‘Everybody AI tomorrow,’” he said. “You’ve just got to be a little bit fleet of foot to be able to dance and learn what you’re being shown and pay attention to what you’re being shown.” He also said that different teams adopt at different speeds, and making room for that allows the learning curve to unfold.
Buesing said he saw the same pattern in his client work. Organizations are now overwhelmingly “in pilot to scale, scaling, or building plans to introduce agentic AI”—but the human side of the equation is lagging the technology. “That wave is coming,” he told Fortune. “And I think organizations may be a little bit slow on that right now.”
The job titles themselves are already in flux. Buesing said he’s already heard companies experimenting with terms like “advocate” or “journey manager” to replace the old “agent” label—partly because it’s become hopelessly ambiguous in the age of AI agents, and partly because the human role genuinely is becoming something new.
“The companies that understand how to unlock this intelligence, engage their people, deploy the tacit knowledge they already have, then use AI are going to win extraordinarily,” he said.
The companies that simply cut, he warned, “will milk the economic value of the knowledge that the AI had from past practice for maybe 10, 15 years. But there’s no more new knowledge being developed because humans develop knowledge, and then the well will run dry.”
The honest answer, as Manos summed it up, is that those six free hours you just saved by using AI aren’t coming anytime soon. What is coming is a widening aperture—more problems to solve, more projects to chase, a bigger version of the job. “The work is not going to go away,” he said. “Pieces and parts of the work may go away, but that just means we’re going to be able to address more.”
Manos noted that Dun & Bradstreet traces its founding to before the Civil War and has survived through business iterations dating back to Abraham Lincoln’s time. The business model of organizing data, he pointed out, used to look very different. “The way they used to do it was, get on a horse, ride into town, figure out who the blacksmith was and who the grocery store was, and then they wrote it down and put it in a book.” The work is the same now as it was then, but all the horses are gone, all the locations are changed. The context has changed, but it still works.
Whether that’s liberation or a treadmill set to a higher speed is shaping up to be the defining labor question of the decade.
