AI in Customer Support: How Companies Save Millions
Recent case studies prove: when AI is implemented correctly, costs drop dramatically — while customer satisfaction rises at the same time.
Customer support is one of the most cost-intensive functions in any company. At the same time, it is one of the most frustrating touchpoints for customers: long wait times, repetitive questions, and opening hours that don't align with users' daily lives.
Since 2023/2024, large and mid-sized companies have been deploying AI-powered chat and voice agents — with measurable, sometimes dramatic results. The following case studies are publicly documented and demonstrate what is possible when implementation is done carefully and with clear objectives.
Industry average 2025/2026: What AI delivers in support
Aggregated data from several hundred companies worldwide
Sources: Gartner 2026, Sobot 2025, Swiftask AI, McKinsey Digital, IBM Annual Report 2025
Klarna: 2.3 Million Conversations in the First Month
Buy-Now-Pay-Later pioneer · 85 million customers worldwide · 35+ languages
The Problem
Klarna operates a global payment platform with 85 million active customers across 45 countries. Thousands of support requests arise daily: refunds, payment disputes, balance enquiries, cancellations. The classic model with human agents reaches its limits at this scale — not because of the quality of staff, but because of the mathematics: 24/7 availability in 35 languages simply cannot be scaled cost-efficiently with human personnel.
A further problem: the average waiting time for a response was over 11 minutes. For customers trying to resolve a payment issue, that is unacceptable. The repeat request rate — customers who contact multiple times because their issue wasn't resolved the first time — was also high.
The Solution
In February 2024, Klarna launched an AI assistant based on OpenAI technology, fully integrated into their CRM systems, order management and transaction databases. The assistant can view transactions, initiate refunds, adjust payment plans and verify accounts — all without human escalation for standard cases.
- Full CRM and transaction database integration
- Natural language processing in 35+ languages
- 24/7 availability without quality degradation
- Intelligent escalation model for complex cases
- Automatic sentiment analysis for prioritisation
Results — First Month
"This AI breakthrough in customer experience puts Klarna miles ahead of competition. We've seen a significant increase in satisfaction while simultaneously cutting costs by tens of millions of dollars."
Key Takeaways
- • Success depended on deep system integration — the agent could actually perform actions, not just respond
- • Customer satisfaction did not decline — it remained at the same level as human support
- • The focus on routine cases (80% of all requests) with a clear escalation path for complex cases was decisive
IBM AskHR — 94% Autonomous Resolution Rate
Global technology corporation · 280,000+ employees · Internal HR automation
The Problem
IBM's HR department received hundreds of thousands of employee enquiries per year — payroll questions, leave requests, benefits queries, policy clarifications. Processing these manually cost millions annually and kept HR professionals occupied with repetitive work instead of strategic tasks. Average response time was measured in hours, not minutes.
The Solution
AskHR — an AI agent trained on all HR documents, policies, processes and historical Q&A pairs. Employees get instant, accurate answers without waiting for an HR advisor. The system integrates with IBM's HRIS (Human Resources Information System) to provide personalised responses based on each employee's specific situation.
- Trained on complete HR policy library and historical FAQs
- Personal context: answers tailored to employee's role, location, tenure
- Integrates with HRIS for real-time data (leave balances, payslips, benefits)
- Multilingual: supports IBM's global workforce
- Continuous learning from resolved cases
Results
Key Takeaways
- • A 94% autonomous resolution rate does not mean 94% of employees are unsatisfied — the opposite: HR advisors now handle only the cases that genuinely require human judgement
- • The investment in high-quality training data (complete policy library, historical Q&A) was the decisive factor for accuracy
- • Internal AI implementations often have faster ROI than customer-facing ones — because usage is predictable and the knowledge base is controlled
What both cases have in common
Deep Integration
The AI agent had access to real systems — not just an FAQ database. It could act, not just respond.
Clear Focus
Neither company tried to automate 100%. They identified the most frequent, clearly definable cases and automated those first.
Escalation as a Feature
The AI solution was designed from day one so that complex or sensitive cases are immediately passed to a human with full context.
Want to achieve something similar for your company?
Talk to us about your use case — we'll show you what's realistic and how long it takes.
Get in touch