Klarna · AI product · Entry · 45 to 60 minutes

Should Klarna scale AI chat support now?

Klarna is a fintech company that lets consumers pay for purchases over time and manages merchant payments. Answer this product management case and get it graded line by line against a weighted rubric.

product judgment
metrics
ai ux
risk management
stakeholder thinking

The situation

You are a PM on Klarna's customer service platform. Leadership is excited about a new GPT-4 powered chat assistant that has shown strong early promise on routine questions like payment dates, order status, and basic account help. The company wants better margins, faster response times, 24, 7 coverage, and support across many languages. Customer service leaders also see a chance to reduce pressure on human agents, who currently spend a lot of time on repetitive tickets. Your VP of Customer Service and a senior product leader have asked for a recommendation: should Klarna move quickly toward AI handling most inbound chat support, or take a more limited rollout? The decision matters because support is a major cost center, but Klarna also operates in financial services, where mistakes can hurt trust and create compliance risk. Some stakeholders want bold automation targets. Others worry that complex billing disputes, fraud concerns, and emotional cases need empathy and careful judgment. You need to propose a product approach that balances efficiency, customer experience, and operational risk.

What to produce

Write a 1 page product recommendation memo, about 600 to 800 words.

Your answer must

State your recommendation on whether Klarna should scale the AI chat assistant broadly now, keep it narrow, or use a hybrid approach, and explain why.
Define the initial user problems and support use cases the AI should handle first, and name at least two cases that should go to a human.
Propose a rollout plan with stages, clear handoff rules to human agents, and what customer controls or disclosures should appear in the chat experience.
Specify the success metrics you would track, including at least one speed metric, one quality or satisfaction metric, and one risk metric.
Identify the main risks, including customer trust and compliance, and give concrete mitigations for each.

Constraints

You cannot assume Klarna can add unlimited headcount, because leadership wants meaningful efficiency gains.
Klarna serves customers in multiple markets and languages, so the solution must work beyond a single country or language.
Because Klarna is in financial services, the assistant cannot give unreliable advice on payments, disputes, or account issues without safeguards.
You are recommending a product decision for chat support only, not a full company reorganization.

Format

Use short section headers and bullets where helpful.

How Klarna actually played it

A short write up on Klarna rolling back AI only support after satisfaction fell. It argues for AI as an assistant with a path to people.

Track satisfaction beside cost
Keep a path to a human
Do not remove people before quality is proven

Built from Klarna's AI Customer Service Experiment by zhuochun.