Companies that rolled AI out fast and cut headcount to match are starting to walk it back, Klarna among the most visible. Data Society’s Catie Maillard sees the same pattern from the People side: rapid AI rollouts are driving real burnout, especially in functions caught between design and execution, and the fix isn’t slowing down out of caution, it’s treating AI adoption like the workplace transition it actually is.. The internet and digital cinema didn’t arrive overnight either. Neither should this.
The department caught in the squish
Maillard’s read on where AI is landing hardest at work isn’t headcount. It’s the people left standing after the change.
“It’s a really hard place to be, and to be fair, another impact I’m seeing of AI is burnout and a lot of feelings of overwhelm in these departments. In the departments that have rapid change, but a lot of times in the departments or functions that are in that squish between design and experience.”
Catie Maillard, Global VP of People, Data Society
The research backs up what she’s describing on the ground. Harvard Business Review found 88% of heavy AI users report increased burnout, and 47% say they aren’t getting enough support to use the tools effectively. Gallup found only 9% of employees feel very comfortable using AI in their role. Fortune reported 39% of workers feel overwhelmed by the changes AI is bringing to their jobs, and Yahoo Finance found roughly 3 in 10 say AI has increased their workload rather than reduced it. None of that reads like a productivity story. It reads like a change management story that skipped the change management.
Move fast and break things doesn’t survive contact with customers
Maillard’s next point is about what happens when companies skip that step entirely, and there’s already a well-documented example.
“I think the only piece that I’ve really been mulling on now is that I think we’re seeing the impacts of companies who moved fast and broke things in the AI world. The impacts are not positive. We’re starting to see companies hire back folks that they let go.”
Catie Maillard, Global VP of People, Data Society
Klarna is the case everyone in HR and customer experience has been watching. In February 2024, the buy-now-pay-later company deployed an AI chatbot it said could do the work of 700 customer service agents, and within a month the bot was handling two-thirds of chats, about 2.3 million conversations, across more than 35 languages. Headcount dropped roughly 40% as the company let attrition do the rest, without backfilling. By 2025, CEO Sebastian Siemiatkowski was telling Bloomberg the AI produced “lower quality” service and that “investing in the quality of human support is the way of the future for us.” Klarna started actively recruiting human customer service reps again, this time explicitly promising customers a person would always be available if they wanted one.
Klarna didn’t get the technology wrong. The bot handled volume exactly as advertised. What the company underestimated was how much of customer service, and by extension most knowledge work, depends on judgment calls a template can’t make, and how fast customers notice the difference.
Other transitions didn’t happen overnight either
Maillard’s conclusion isn’t to slow-walk AI out of fear. It’s to manage the rollout the way past technology transitions actually got managed, which was never as fast as the retelling suggests.
“My takeaway here, long term, is that we need to be slow and thoughtful in this transition. We need to really manage this like it is a workplace transition, similar to how we managed the rollout of the internet, similar to how we managed the switch from film to digital. That did not happen overnight, and I think we need to stop expecting this to happen overnight, because it’s overwhelming brains, and it’s just going to cause this to meet more resistance than it needs to.”
Catie Maillard, Global VP of People, Data Society
Hollywood’s own switch from film to digital projection is a fair comparison. The first digital cinema systems went into theaters around 1999 and 2000, and the industry didn’t cross the finish line on nationwide conversion until roughly 2013 to 2015, over a decade of theaters upgrading projector by projector, market by market. The internet’s spread through households and workplaces played out over a similarly long stretch, not a single rollout event. Neither transition was gentle, and neither was instant. Both still took years longer than most companies are currently giving AI.
Frequently Asked Questions
Klarna deployed an AI chatbot in early 2024 that handled two-thirds of customer chats and let headcount shrink roughly 40% through attrition. By 2025, the company reversed course, citing lower service quality, and began actively rehiring human customer service representatives.
Research suggests yes. Harvard Business Review found 88% of heavy AI users report increased burnout and that 47% feel unsupported using the tools, while Fortune found 39% of workers feel overwhelmed by AI-driven change in their jobs.
Functions in the middle of rapid change with unclear ownership, particularly roles that sit between design and execution, appear especially exposed to overwhelm, since they absorb both the tooling changes and the process changes at once.
Longer than most people remember. Hollywood’s conversion from film to digital projection ran from roughly 1999 to the mid-2010s, more than a decade. Household and workplace internet adoption unfolded over a similarly extended period.
Treat it as a managed workplace transition, not a single launch: pace the rollout, support the teams absorbing the most change, and expect resistance to shrink as the timeline realistically lengthens rather than trying to compress it.
Next steps
If your AI rollout is measured in weeks instead of years, that’s the first thing to slow down, not the technology itself. Data Society’s AI Advisory work helps leaders pace a rollout like the transition it actually is, and AI upskilling programs support the teams absorbing the most change.
