Accenture is now tracking how often its executives log in to AI tools — weekly, by name — and tying the results to promotions. The Financial Times calls it a "carrot and stick" approach: senior staff are less willing to adopt AI than juniors, and the consulting firm has decided the answer is measurement and enforcement. Promotions now require "regular adoption." This is what a mandate looks like. It is worth asking what the evidence behind it looks like.
Seven Hundred
In February 2024, Klarna announced that its ChatGPT-powered AI chatbot was "doing the equivalent work of 700 full-time human agents." It was the most cited data point in the corporate AI narrative — proof that AI could replace an entire customer service operation and come out ahead.
-
FEB 2024Klarna claims AI chatbot is doing the work of 700 full-time agents. The number becomes the poster child for AI-driven workforce reduction.
- May 2025 Klarna starts hiring back human customer service workers after the AI approach led to "lower quality."
- Sep 2025 Klarna IPOs on NYSE at $45.82. Sequoia gains ~$2.65B.
- Nov 2025 CEO Siemiatkowski says he is "nervous about the size of these investments" in AI data centers.
-
FEB 2026Klarna reports Q4 revenue up 38% but a $26M net loss — down from a $40M profit a year earlier. Stock is down 66% since IPO.
The arc: replaced 700 workers with AI, discovered the quality was worse, hired humans back, IPOed on the AI narrative anyway, and crashed. The number that launched a thousand corporate mandates turned out to be a warning, not a proof point.
The Intensification
An eight-month study published in Harvard Business Review this month found that AI tools at a US tech company didn't reduce work — they intensified it. Researchers spent two days a week embedded with roughly 200 employees from April to December 2025 and found three patterns: employees expanded into tasks they'd previously outsourced, blurred work-life boundaries because prompting AI felt like chatting rather than working, and managed multiple active threads simultaneously.
You had thought that maybe, because you could be more productive with AI, then you save some time, you can work less. But then really, you don't work less. You just work the same amount or even more.
The study's conclusion is precise: initial productivity gains gave way to workload creep, cognitive fatigue, burnout, and weakened decision-making. Workers felt more productive. They were not less busy. The promise was efficiency. The delivery was intensification.
Four Percent
A study of 12,000 EU companies published the same week quantified what AI actually delivers: a 4% average increase in labor productivity, with no evidence of reduced employment in the short run.
Four percent is real. It is also not the transformation that justifies tracking executives' login frequency, tying promotions to tool adoption, or replacing 700 workers. It is the kind of gain that would not survive a quarterly earnings call as a headline number. It would be a footnote.
Bloomberg noted a year ago that AI adoption had outpaced PCs and the internet, but "evidence of its boost to productivity is thin on the ground." Gallup measured in mid-2025 that 25% of US workers use AI weekly. Adoption is accelerating. The evidence beneath it has not kept pace.
The Infrastructure
On the same day as the Accenture story, Amazon dethroned Walmart as the world's largest company by revenue — $717 billion in 2025 sales. The milestone belongs mostly to AWS, the infrastructure layer that powers the AI tools everyone else is mandating. The company selling the shovels just became the world's biggest business.
The mandate flows downhill. Investors need the narrative. CEOs tell it. Consultants like Accenture enforce it. Employees get their logins tracked. And the infrastructure providers — Amazon, Microsoft, Google — collect the margin at every step. The adoption is real. The four percent is real. The transformation is a mandate looking for its evidence.
The companies tracking AI adoption are not the same companies proving it works.