Situation — AI and sovereignty

Our AI POC works in demos, but nobody dares take it live.

Context

The prototype impressed everyone in the meeting. It answers, it sorts, it drafts. Then it was left to gather dust.

Nobody wants to sign off on going live. What if it gets things wrong in front of a client? Who monitors it? Who keeps it running once the project team has moved on?

The problem

  • A dormant POC is an investment that earns nothing.
  • Taken live without safeguards, it can lower the quality your clients experience.
  • With nobody to look after it, the tool degrades and nobody trusts it any more.

Case study

Real-life caseMay 2025

Klarna, fintech, Sweden

In 2024, Swedish fintech Klarna claimed its AI assistant was doing the work of 700 agents. At the time, it handled 75% of customer conversations, in more than 35 languages.

Customers complained about the assistant's limits. In May 2025, the CEO admitted that cost had weighed too heavily in the decision and that quality had dropped.

Klarna started hiring human agents again, working remotely, so that its customers are guaranteed a human contact.

Read the sourceKlarna Dials Back its AI Customer Service Strategy — Now It's Hiring Humans Again — Maginative (opens in a new tab)

Why us?

  • We make your POC robust and take it all the way to production, as soon as you give the go-ahead.
  • We test its answers, including attempts to misuse it, before it talks to your clients.
  • Security, monitoring and documentation: you see what the tool does, and you can correct it.
  • We train the team that keeps it running.

Contact us

Talk to an expert

A free 30-minute video call with one of our experts, then your free quote, with no commitment.

Take the diagnostic

A few questions, the first already answered for your situation: your answers help us prepare the video call. Free, no sign-up, about 2 minutes.