(This column originally appeared in Forbes)
One of my favorite AI stories this year is the reckoning of Swedish Buy-Now-Pay-Later platform Klarna.
“Klarna’s AI assistant handles two-thirds of customer service chats in its first month!” the company giddily announced to the public in early 2024, further trumpeting that it did the work of “700 customer service agents.”
And that wasn’t all.
“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,” the company’s CEO said last September. “By simply not hiring, which we haven’t done since September … the company is kind of becoming smaller and smaller.”
Replacing people with AI. Great idea! Except…not.
Just a few weeks ago Klarna’s CEO — who literally used an AI doppelganger to deliver an earnings report — admitted to Bloomberg that the company is “slowing down its job cuts” and getting back to hiring real people to do customer service because the company discovered that people actually want to talk to people.
Separately, a representative from Klarna said that the company is “not reversing on AI” and is “continuing to invest heavily in AI, including rebuilding our tech stack to be AI-first.” But the company also admitted that their CEO “acknowledged that an overemphasis on cost — not AI itself — led to lower quality.”
Turn’s out Klarna’s not the only company experiencing buyer’s remorse from AI.
For example, a company that announced “driverless trucks” built on an AI platform that would be rolling through Texas recently discovered that they actually needed drivers after all. Translation company Duolingo aspired to replacing their translators with AI bots only to change course after a backlash from its community. Etsy has faced similar backlash from its community thanks to non-human-looking-AI-generated products flooding its system.
After firing 8,000 employees in order to replace them with AI systems, IBM wound up re-hiring many due to “gaps in service, dips in employee morale, and delays in resolution.” Australian telecom company Optus, after attempting to automate functions with AI, turned around once it found that AI couldn’t handle many issues as well as humans. McDonalds ended its drive-through AI-based ordering test after discovering that humans could do a better job servicing their customers.
Which is why it should be new surprise that a recent report by organizational design and planning platform firm Orgvue found that as much as 55 percent of companies that had laid off staff due to AI automation now regret the decision. According to the report “these companies found that AI could not fully replicate the nuanced understanding and adaptability of human workers, leading to a reevaluation of their workforce strategies and a renewed emphasis on human roles.”
Uh…duh?
This stuff just isn’t ready for prime time in 2025. And companies who think that they can use AI instead of humans this year are going to learn the same lessons as Klarna, McDonalds and others.
For starters, people aren’t stupid.
I stayed at a Marriot conference resort recently and called to find out if the restaurant was still open. A very human-like bot answered my call, and at first I thought it was a real person. But I figured out quickly that it wasn’t. Something was just…off.
Just like you can sense someone else in a room with your eyes closed, humans innately know when they’re not dealing with another human. We quickly figure out when we’re talking to a bot. Maybe not right away. But give me just 60 seconds on the phone with an automated, AI-driven voice and I’ll be able to figure out that the “person” on the other end isn’t a person. AI isn’t yet capable to respond quickly, jump back and forth between questions and issues and generally behave like a human. All you have to do is ask a personal question or veer off track from the main conversation and most AI bots get confused and then sound like…a bot.
Next, people will want to talk to humans…sometimes.
How many times do you furiously punch “0” on the phone the minute an automated system picks up?
Navigating through these help desk mazes can be infuriating and time consuming. But that doesn’t mean there isn’t a role for AI. A good customer service system using AI will eliminate the “press 1 for customer service, press 2 for sales, etc.” with a human-like voice asking questions and (hopefully) giving answers quickly. But therein lies the most important thing: if an answer can’t be gotten quickly or if the caller simply wants to vent to a human, there has to be quick way to take the red pill (or is it the blue pill?) and get out of the Matrix.
My advice to companies is to understand that people do want to talk to people sometimes and to understand that not every question is a simple question. Be transparent when a bot answers the phone and give us a way out. Don’t get rid of your customer service staff. Use AI to screen and answer low-level questions and then use your humans for everything else.
“Humans deeply care about what other humans think — it’s something that seems hardwired into us,” OpenAI’s CEO Sam Altman said last year at a conference. “While we may keep developing better tools, our focus will always return to one another.”
Finally, the companies that advertise their internal AI initiatives are making a mistake.
Many companies think they’re doing the world a favor by issuing press releases advertising their AI investments. My advice: don’t. It’s simply bad PR. Klarna made the mistake of gleefully telling the world how they’re literally replacing humans with AI. Now they look dumb.
Please don’t advertise how you’re replacing people with AI. Don’t be like the Shopify CEO who publicly stated that it has to be proven that any new job can’t be done with AI. He looks like a villain. No one wants to hear it. Keep it to yourself.
Many workers are terrified of losing their jobs and when a company like Shopify does this kind of stuff it says to them: we don’t care about you. Good luck finding talent after that. Who wants to work for a company that considers their people so value-less that they’ll not only replace them with bots but tell the universe how smart they are by doing so?
AI is great and it will get better. But never believe that it will be good enough to fully replace a human interaction. It never will.
