As Klarna flips from AI-first to hiring people again, a new landmark survey reveals most AI projects fail to deliver

I guess it wasn’t possible to squeeze more investor money out of the AI will take all the job narratives.

This was THE ChatGPT will replace customer service flagship project with OpenAI using Klarna as a GTM partner [2] and blasting their integration case study to the world to smokescreen enterprises into thinking the technology is production ready.

It is not.

Prompt injection remains largely unsolved, meaning agent based systems beyond traditional workflows are vulnerable to client shenanigans and can not be given authority over decisions like refunds.
Hallucinations continue to be unsolved and, by our understanding of science, can likely never be eliminated. Even if hallucination rates were decreasing, they are still so far away from triple-9 reliability needed for effective automation that only costly mitigation scaffolding can make them somewhat usable. And in regulated industries like finance, where money is on the line, that’s just not good enough.
And no, RAG does not solve either (although I’ve had plenty of systems integrators trying to pitch me just that..)

The only systems that “work” end up being so constrained and locked down that one instinctively has to as “Have you tried just giving the user a decent search experience on your website perhaps, it’s cheaper and better?”

As Klarna flips from AI-first to hiring people again, a new landmark survey reveals most AI projects fail to deliver

C4AIL Commentary