.jpg)
In the final season of Mad Men, Peggy Olson throws out a perfectly approved pitch and starts over. She returns to a blank legal pad. She revisits a restaurant she'd already been to more than a dozen times. Nothing about her process is efficient — and that's precisely why it leads to something better than what she had.
It's a useful frame for thinking about how artificial intelligence is reshaping the modern workplace. The promise is significant: AI can eliminate friction, accelerate output, and free workers for more meaningful challenges. But the experience of many workers — and a growing body of research — suggests the picture is more complicated.
The Productivity Promise Meets a Complicated Reality
Business leaders have made ambitious claims about what AI can do for work. Mark Cuban has argued that AI empowers creators to become exponentially more creative. Bill Gates, Eric Yuan of Zoom, and JPMorgan Chase CEO Jamie Dimon have each suggested that widespread AI adoption could eventually free white-collar workers from the traditional five-day workweek.
The data, so far, tells a more measured story. MIT research found that 95% of AI pilot programs had not produced a measurable return on investment. A Harvard Business Review study published this February, which followed 200 employees at a U.S. technology company over eight months, found that AI tools didn't reduce workloads — they intensified them. Workers completed tasks more quickly but took on longer hours and expanded their responsibilities. Many began using AI during breaks, eroding the boundary between work time and rest. The researchers flagged risks of cognitive fatigue, burnout, and weakened decision-making.
The Hidden Value in Difficult Work
One of the subtler risks of AI adoption at work is the temptation to use it to bypass cognitively demanding tasks rather than support them. Cal Newport, a computer science professor at Georgetown University and author of Slow Productivity, describes this pattern clearly: much current AI use, he argues, is less about removing administrative tasks that stand in the way of meaningful work and more about helping workers "reduce or avoid the peak cognitive strain of doing harder thinking."
The blank page, rather than being engaged with, gets routed around.
Ben Armstrong, executive director of MIT's Industrial Performance Center, raises a related concern about what gets lost when routine tasks are fully automated. Manually reviewing a data set, for instance, is tedious — but it also surfaces context, gaps, and anomalies that a polished AI summary might quietly obscure. "I worry," Armstrong says, "that maybe we're not as good at the higher value-added work if we don't do some of those mundane tasks."
Research supports the idea that routine work carries its own cognitive benefits. A University of Central Lancashire study found that completing boring tasks can prompt new bursts of creative problem-solving. There is often something useful hiding in the tedium.
The Risk of Outsourcing Creative Thinking
The concern extends to creative work specifically. An MIT study found that people who regularly used ChatGPT to assist with writing became more dependent on the tool over time and consistently underperformed compared to those who worked without AI assistance. Research from the Wharton School found that while AI tools can enhance an individual's output, they tend to flatten creative diversity across groups. A Columbia Business School study found that large language models exhibit a bias toward whichever option is presented first — a structural tendency that could undermine the kind of iterative, evolving thinking that leads to genuinely original ideas.
Emily DeJeu, a professor at Carnegie Mellon University's Tepper School of Business, frames it this way: work that requires holding multiple ideas simultaneously, synthesizing them, and producing something new is "cognitively very taxing." The idea that AI can simply assist with or accelerate that process, she says, "is a bit fallacious."
Some workers have arrived at similar conclusions on their own. Karim Adib, a public relations manager at Search Atlas, experimented with using generative AI for brainstorming before returning to a more analog approach — stepping away from his desk with a notebook, going for a walk, visiting a library. His reasoning was pragmatic: "Everyone else has access to ChatGPT as well. If someone else and I get the same idea, executing in the exact same way, that gives me zero advantage."
The Deeper Source of Workplace Fatigue
Newport is careful to note that cognitive strain and blank pages are not, themselves, the primary drivers of workplace burnout. The more significant culprit is what he calls "frenetic" communication — the constant interruption of messages, notifications, and meetings. A 2025 Microsoft study found that workers using its suite of products were interrupted by chats, emails, and meetings an average of 275 times per day. The early adoption of AI agents and chatbots, Newport argues, risks compounding this problem rather than solving it.
"A lot of the things that make us so busy and drive up our long work days," he says, "it's not really problem-solvable by AI. It's a problem solvable by workplace culture."
A More Sustainable Approach
None of this is an argument against AI in the workplace. The tools are genuinely useful, and when thoughtfully applied, they can free up meaningful time and capacity. But organizations that want to build sustainable, high-performing cultures need to be deliberate about what they automate and what they protect.
Turning nothing into something — sitting with a problem, working through false starts, arriving at an idea that is genuinely your own — is how people develop judgment, build confidence, and produce work they feel ownership over. A few quiet, unoptimized moments in the workday are not inefficiencies to be eliminated. Often, they are where the real work happens.