World's largest call center company. AI agents replacing humans.
Teleperformance employed 500,000+ people across 170 countries, operating call centers for the world's biggest brands. When Klarna announced its OpenAI-powered chatbot handled 2.3 million chats in one month — equivalent to 700 agents — it became the poster child for AI disruption in customer support. The stock crashed 27% in a single day. JPMorgan warned that 'automation-related deflation is only set to accelerate.' Short interest rose to 6.4% as bears called it 'one of AI's first victims.' The per-seat billing model that powered decades of BPO growth is structurally broken.
AI-powered customer service agents handle millions of conversations at a fraction of the cost. Klarna's OpenAI chatbot replaced 700 full-time agents in its first month. Companies are deploying autonomous AI agents that resolve 80%+ of support tickets without humans.
Peak: €410/share, 500K+ employees, dominant global BPO
Early AI fears begin eroding stock; chatbot pilots proliferate
Klarna AI chatbot replaces 700 agents; stock crashes 27% in one day
JPMorgan warns of accelerating automation deflation; stock hits 7-year lows
Stock -75% from ATH, short interest at 6.4%, mass contract renegotiations
Deploy AI customer service agents that handle 80%+ of support tickets autonomously. Route only complex, emotional, or escalated cases to humans. One AI agent replaces 50-100 human agents for routine inquiries.
Audit your current support tickets: categorize by complexity and resolution type
Identify the 80% of tickets that are routine (FAQs, status checks, simple requests)
Build an AI agent with your knowledge base, past tickets, and brand voice guidelines
Set up human handoff rules: sentiment thresholds, topic exclusions, VIP routing
Deploy on chat first, then expand to email and voice channels
Monitor resolution rate, CSAT, and escalation patterns weekly
You are a customer support agent for {{company}}. Your knowledge base is below: {{knowledge_base}} Rules: - Be friendly, concise, and helpful - If you can resolve the issue from the knowledge base, do so - If the customer is frustrated or the issue is complex, say: 'Let me connect you with a specialist who can help further.' - Never make up information about policies, pricing, or product features - Always confirm the resolution before closing Customer message: {{message}}
Categorize this support ticket into one of these categories: {{categories}}. Also assess: - Urgency: LOW / MEDIUM / HIGH / CRITICAL - Sentiment: POSITIVE / NEUTRAL / FRUSTRATED / ANGRY - Can AI resolve: YES / NO (explain why not) Ticket: {{ticket_content}} Respond as JSON.
I have {{count}} resolved support tickets from the past 6 months. Analyze them and create a structured knowledge base with: 1. Top 20 most common questions with ideal answers 2. Decision trees for multi-step troubleshooting flows 3. Escalation criteria (when AI should hand off to human) 4. Response templates for each category 5. Edge cases that require human judgment Format as a structured document the AI agent can reference.
Continuously improve: feed resolved tickets back into the AI knowledge base