Analysis. With AI – and especially GenAI – there are almost no limits as to how it can enhance efficiency and replace human work force. But with AI – especially chatbots in customer services – companies should ask themselves a lot of questions before implemetation.
Klarna, the huge financial financial tech giant, is probably the best case to date showing that we have to stop hyping the efficency promises of AI. In March 2024, Klarna sends out a press release claiming that their AI chatbot based on OpenAI’s GPT does the same work as 700 fulltime staff. In May the same year, the Wall Street Journal reports that Klarna had cut sales and marketing spending the first quarter by 11% and that 37% of that was due to AI. Then in August, Klarna’s CEO Sebastian Siemiatkowski told Reuters that; “About 12 months ago, we would have been about 5,000 active positions within the company, and we are now down to about 3,800.” It was done with the help of AI without any layoffs but because they’d stopped hiring.
So far so good. Then, two American news outlets looked into the claims. TechChrunch found 50 job postings by Klarna in beginning of 2025, and i February, the New York Times interviewed Mr. Siemiatkowski, who admitted that he had been exaggerating.
Why did the Klarna CEO exaggerate and brag about laying off humans – probably annoying a lot of his human staff?
May be because Klarna is preparing to IPO in 2026 and he tried to attract Wall Street’s attention with a compelling narrative, as Alexander Chukovski writes. May be he is just so obssessed with AI as many others – especially the masculine side of humans. The industry – with huge economic interests – certainly has blown up the prospects of generative AI. But Mr. Siemiatkowski from Klarna seems so fascinated about it that he has gone as far as to cloning himself into an avatar talking to his staff, investors and everybody else on LinkedIn.
There is no shortage of promises that AI will save companies and the public sector a lot of money. One of the latest promises is from a report made by Boston Consulting Group on behalf of the Danish Employers Association ‘GenAI – a Significant Potential for The Danish Public Sector 2040.’ It estimates – in big tech lingo – that GenAI will revolutionize the public sector:
“GenAI can provide a productivity boost in the range of 48-55 billion kroner in the public sector in 2040. This means that AI can help free 84-000-92.000 full time positions in 2024, which is the equivalent of 10-12% of the total public sector workforce.“
In the report, following examples of potential efficiency gains in the public sector are:
Build models that can predict demand. Give better service with better analysis and feedback. Coordinate routes in public transport. Offer support 24/7 with chatbots. Personalize services and communication.
Optimize schedules and summarize legal documents, code and programming, summarize meetings, help staff make decisions, make better documentation.
Make new political programs and legal analysis. Evaluate political alternatives and simulate the consequences of a political move. Check implication of new politics and generate drafts for speeches.
Some research show that AI can result in efficiency gains. According the The Financial Times, for example, Lenovo claims that their programmers and engineers have become faster and better in programming by 10 per cent.
A recent paper by researchers at Microsoft and three universities found that programmers’ use of an A.I. coding assistant called Copilot, which proposes snippets of code that they can accept or reject, increased a key measure of output more than 25 percent.
And according to Antropic, who is behind the chatbot Claude and thus has vested economic interests (like Microsoft) in the efficiency promises, 50 per cent of all whitecollar jobs could be replaced by AI within a 1-5 years. Also here we’re talking abou coding, according to the podcast Hardfork.
Other research, however, show that the gains are exaggerated: Data from 2024 showed efficency gains by only 2,8% on average by using AI chatbots, according to University of Chicago and Copenhagen University.
Human and Time vs. Chatbots and Efficiency
Now, talking about chatbots, research shows that most humans don’t really want to talk to a machine. 70% of us say that we’d rather be adviced by a human than a chatbot according to the Danish media outlet, Finanswatch, who asked the analysis company Wilke to do the research on the use of AI in the financial sector. 37% are worried or very worried about the use of AI. And around one third will be okay using a chatbot, if they can get an answer outside opening hours.
In the discussion about the effects of AI, there is unfortunately only one success criteria and that is profits/cost savings. AI adoption is fast. According to Eurostat, Denmark ranks highest with over 27% of businesses using at least one AI technology in 2024—twice the EU average of 13.5%. And when asked what kind of AI tech they use, it is mostly in sales and marketing – thus chatbots. Of course the adoption is much faster in the US.
There is no doubt that AI – both generative AI and in particular machine learning – is already optimizing parts of all sectors. Especially behind the scenes, where you are not in direct contact with customers or citizens. Customer service staff or civil servants can of course use AI to help them give better service by for example finding the right information or law very fast.
But talking to a chatbot is so annoying that I think we will see a revolt sooner of later. Why do the many services installing chatbots not ask themselves what their customers actually want. There are of course exceptions; a few companies are sticking to human customer support. But the majority have or will chose AI because of the efficiency promise.
Before implementing AI, following questions scould be considered by both the private and especially the sector:
Who will be in control or data – store them and have access to them
What societal and cultural values will affect the results coming out of the AI service
Will we be able to control the AI system and do we need that control
Will it be fun for our human staff working with us having to cooperate – or compete – with more and more machines
Will we be able to retrain and reeducate our staff whom we replace by AI
Will our staff be able to fact check the results delivered by generative AI
Will our staff be able to challenge the results form the machine, if they have a different opinion or gut feeling
What are the efficiency gains worth compared to the loss of jobs for our staff
What is the effect or our AI use on the environment
Do we care about the copyright issue when using generative AI
Will we lose customers, if we employ a chatbot in customer service or
Will we gain new customers if we keep humans in customer service (and the reception)
Will the use of the AI service help Europe obtain digital sovereignty
Do send me more good questions to ask when implementing AI-services to info@dataethics.eu
This is 100% human-made. Not GenAI is used.
