3 documented workforce events with varying AI attribution strength. This is an evidence index, not a finding that AI directly caused every action.
Numeric event figures, excluding role equivalents; not unique people.
Latest verification among these records.
The total sums reported or estimated numeric headcount attached to these events and excludes role-equivalent claims. It can still include future plans, partial actions, and overlapping rounds, so it must not be read as a unique-person total or as jobs caused solely by AI.
Open each event for the original quote, source links, and attribution qualifications.
Reported scope: Reversal of prior 700-agent AI replacement claim; staged rehire of human support
Klarna publicly walked back its all-AI customer service strategy, with CEO Sebastian Siemiatkowski admitting the company 'went too far' and beginning to rehire human agents under a flexible (Uber-like) model. This is the canonical disputed/reversed AI-replacement case.
“We went too far.”Evidence: MLQ.ai ↗
Reported scope: ≈40% workforce reduction since 2022 (from ~5,000 to ~3,000) attributed in part to AI
CEO Sebastian Siemiatkowski told CNBC that Klarna had shrunk its workforce by ~40% since 2022, in part by deploying AI for customer service. Klarna later partially walked back the AI-only customer-service strategy (see klarna-2026-reversal); flagged as disputed because of subsequent rehiring.
“AI is being used to do the work that was previously done by 700 customer service agents.”Evidence: CNBC ↗
Reported scope: Equivalent of 700 full-time customer service agents
Klarna and OpenAI announced an AI assistant doing the work of 700 full-time customer service agents in its first month, handling two-thirds of customer chats. CEO Sebastian Siemiatkowski projected $40M in annual profit improvement. The company did not disclose direct layoffs at announcement but had reduced headcount via attrition and prior layoffs.
“It is doing the equivalent work of 700 full-time agents.”Evidence: PR Newswire (Klarna) ↗
Return to company timelines, browse the full tracker, or filter documented actions by 2026 event date.
It sums reported or estimated headcount attached to tracked events, excluding role-equivalent claims. It does not represent unique people because rounds may overlap or include plans and estimates.
No. It groups events for one company. Each entry's attribution type, strength, confidence tier, evidence notes, and source links state the specific supported claim.