Sponsored by
Table of Contents
In 2023, Klarna replaced 700 customer service workers with an AI chatbot.
The CEO announced it publicly and often. The chatbot handled two-thirds of all customer conversations. It matched the productivity of the 700 agents it replaced. By 2025 Klarna went public and the AI efficiency story was central to the IPO. Shares surged 30% on debut.
By September 2025 Klarna was quietly rehiring humans.
The CEO told Bloomberg: "We focused too much on efficiency and cost. The result was lower quality, and that is not sustainable." He said this ten months after going to market on the exact opposite thesis.
Klarna is not an outlier. It is the leading edge of a pattern that the data is now confirming at scale.
Forrester's 2026 Future of Work report found that 55% of employers regret replacing workers with AI. For context, fewer people regret their tattoos. A Gartner survey of 321 customer service leaders found that only 20% had actually reduced staffing because of AI. The vast majority kept headcount steady and used AI to handle increased volume alongside existing teams. Gartner predicts that by 2027, 50% of companies that cut customer service staff for AI will rehire people to perform similar functions.
The replacements are becoming un-replacements.
The Pizza Hut case makes the mechanism visible.
Yum Brands acquired an AI delivery management system called DragonTail in 2021 for $72 million and mandated it across its franchise network. One franchise operator, running 111 Pizza Hut locations across the northeast United States, had previously delivered more than 90% of orders within 30 minutes. After DragonTail was deployed that number fell to approximately 50%.
The AI did not malfunction. The algorithm sequenced orders correctly. What it did was change the information environment for DoorDash drivers, who are independent gig workers optimizing for their own earnings. DragonTail gave drivers real-time visibility into kitchen operations, including tip amounts and order timing. Drivers began waiting inside stores to batch multiple high-tip deliveries together. Finished pizzas sat on racks for up to 20 minutes before pickup.
Two independently optimizing systems, DragonTail's algorithm and DoorDash's gig marketplace, were connected without anyone modeling how they would interact. The result was an outcome neither system was designed to produce. The franchise is now suing for $100 million in damages.
The AI worked exactly as specified. The business decision to deploy it without understanding the system it was entering did not.
The World's Biggest Dev Event Hits Silicon Valley
WeAreDevelopers World Congress comes to San José, CA — September 23–25, 2026. 10,000+ developers, 500+ speakers, and the full software development lifecycle under one roof, in the heart of Silicon Valley.
Kelsey Hightower. Thomas Dohmke (fmr. CEO, GitHub). Christine Yen (CEO, Honeycomb). Mathias Biilmann (CEO, Netlify). Olivier Pomel (CEO, Datadog). The people actually building the tools you use every day — all on one stage.
AI, cloud, DevOps, security, architecture, and everything real builders ship with. Workshops, masterclasses, and the official congress party.
The pattern underneath Klarna and Pizza Hut is identical. In both cases the AI was technically functional.
In both cases the failure was the decision to use AI as a wholesale replacement for human judgment rather than an extension of it. In both cases nobody built the parallel systems, the gradual testing, the redundancy that would have revealed the problem before it became expensive.
The companies seeing returns from AI are doing something differently. Unity saved $1.3 million annually by using AI to summarize support ticket histories without cutting a single agent. Insurance teams reduced claims processing time from an hour to ten minutes by using AI to compile documentation for human adjusters who still made the final decision. Logistics carriers using route optimization alongside human dispatch supervisors reported 18% fewer delivery delays because the technology handled the calculation and the humans handled the local knowledge.
A May 2026 Gartner survey of 350 global executives found that workforce reductions had no correlation with return on investment from AI. What did correlate with higher ROI was what Gartner calls people amplification. Companies that used AI to make existing employees more productive, investing in skills and roles that let humans guide and scale the technology.
The organizations improving returns were not the ones that eliminated the need for people. They were the ones that amplified them.
For builders, the signal in this data is specific.
The market for AI replacement tooling is contracting. The market for AI augmentation tooling, products that make existing human judgment faster, more accurate, and better informed, is expanding. The companies that went hardest on replacement are now publicly reversing course. The procurement conversations that follow those reversals will not be about replacing the humans they just rehired. They will be about making those humans more effective.
That is a different product. It is also a larger and more durable one.
404 Found covers AI developments from a European Insider, three times a week.
Renewals stop being a fire drill.
Most churn blindsides the CSM in renewal week. Champion left. Usage dropped. NPS slid months ago.
A colleague in Slack watches the signals around the clock. Your CSMs catch every risk months before renewal.
11,000+ teams use Viktor daily. SOC 2 certified. Your data never trains models.