Billuminate
What we find when we measure how financial companies actually answer their customers. Industry reports, and the thinking behind them.
In the September issue of Kreditnytt, the newsletter of the Swedish Credit Association (Svenska Kreditföreningen), Billuminate founder and CPTO Tim Bisander presents our study of Swedish consumer lenders’ customer chats. Only twelve of 31 lenders could be asked a question while logged out, four questions in ten went unanswered, and roughly one answer in five was wrong.
ReadFive of the twelve measured lending chats answer in Swedish only, and seven manage English. But the larger half of the problem is harder to see: a model trained on an ocean of English and a trickle of Swedish learns the pattern without the details, and fills the gaps from someone else's rules.
ReadSince 2 August 2026 the EU AI Act requires that a person interacting directly with an AI system is informed of that fact. Three weeks later we looked at how twelve Swedish consumer lending chats handle it, and why it matters when one answer in five is wrong.
ReadThirty-one Swedish consumer lenders, measured from the outside. Forty-five per cent have no chat at all; those that do failed to answer 44 per cent of the questions put to them, and nearly one answer in five was wrong.
ReadMost chatbots built on large language models will give your customer a faulty answer with exactly the same confidence as a correct one. Looking at the answer, you cannot tell which one you just received. The tone, the flow and the wording are the same.
Read
Demo
Powered by Billuminate
