Feature
Klarna bragged that its chatbot handled 2.3 million customer chats a month. Unfortunately, the bill for lost trust is still climbing.
It looked like a masterstroke: slash headcount by 24% (roughly 1,200 people), let AI handle the load and bask in headlines about innovation and efficiency. CEO Sebastian Siemiatkowski even boasted, “AI can already do all of the jobs that we as humans can do … we stopped hiring about a year ago."
For a while, the metrics agreed. Three-quarters of customer interactions were routed through bots. Resolution times improved. Labor costs dropped. The boardroom loved it.
The numbers weren't telling the full story.
Customers complained about robotic responses, inflexible scripts and the Kafkaesque loop of repeating their issue to a human after the bot failed. Sure, the AI could handle questions. But it couldn’t handle nuance, refunds or loyalty.
Klarna reversed course. The company is now rehiring humans to restore the empathy it amputated. The hiring freeze is over. The cleanup has begun, with Siemiatkowski sheepishly admitting, "From a brand perspective, a company perspective, I just think it’s so critical that you are clear to your customer that there will be always a human if you want."
Klarna isn’t alone. It’s just loud enough to make the problem obvious.
The AI ROI Mirage
AI‑driven staff reductions look good on flashy PowerPoint slides. Labor costs vanish. Productivity graphs spike. Someone, probably the person who thinks that AI can’t replace them, gets a bonus.
But look deeper. The costs don’t disappear. They simply look different.
Customer churn eats revenue as frustrated users bolt, and winning a replacement buyer can cost up to seven times more than keeping the one you lost. Klarna’s own chief conceded quality slipped. One viral complaint, and the chatbot’s efficiency is irrelevant. Great, you saved $50,000 on payroll, but you lost $5 million in lifetime value.
Rehire whiplash kicks in when chatbots fumble and firms scramble to re‑staff. Research shows that Klarna is in the majority: 55% of companies that executed AI-driven layoffs now regret it. The “obsolete” job title returns as “mission‑critical,” all with the addition of recruiter fees, signing bonuses and lost trust.
Reputation damage hits next, just ask Air Canada. The Canadian-based airline was held liable after its AI tool misquoted refund policy. The tribunal ordered compensation and the story went viral as a lesson in what not to do. Your clever little chatbot can mishandle policy at an incredible scale.
Tech write‑offs follow quickly, with 42% of enterprises scrapping most AI projects last year and seven out of 10 generative‑AI deployments missing ROI targets entirely. The high water mark to date remains MD Anderson's 2017 $62 million IBM Watson Oncology write‑off, an embarrassing and costly failure.
Operational drag finishes the picture. McDonald's yanked AI drive‑throughs after viral misfires such as bacon on a sundae, or a $200 nugget spree. These stories might be a little funny. But there’s a difference between being funny once and being funny when your service pipeline melts and managers have to explain why a burger arrived with 10 patties.
Klarna is reversing course, but it’s far from the only company feeling the burn from rushing too quickly into AI. The cost to fix it will be expensive and probably not accounted for when they made the cuts.
Treat AI Like a Scalpel, Not a Cudgel
We can’t put our heads in the sand about AI, but it’s time to stop thinking AI-first. That mindset guarantees that every project, no matter how square, gets the round peg of AI.
Instead, we should begin every project or opportunity by asking what success looks like. Define the metrics that matter (i.e. retention, conversion, share of wallet) and what you don’t want to lose (i.e. customers, employees, trust) before reaching for code. If a tool can move that metric without negatively impacting what you can’t lose, give it a clearly fenced role. If it can’t, walk away.
A healthier approach treats AI like a scalpel for precision work. Feed it the rote, the repetitive, the pattern‑heavy tasks. Humans stay on the messy edge cases where nuance, empathy and judgment protect the brand, employees and customers.
Prototype in daylight instead of in the shadows. Run a tight pilot, track live quality metrics and give every customer a red button that routes straight to a human. That escape hatch is an insurance policy against the next viral fiasco.
A smarter cadence is a risk management tool. Moving slightly slower costs less than a six‑month rollback. When you do need to act fast, run the full ledger on visible costs and hidden ones before your reduction in headcount causes the lifetime value you’ve spent so much to build to vanish.
AI only cuts costs if you count all the costs.
Trust Still Matters
AI can trim payroll in a quarter. It can gut value just as fast.
Conference
Oct
20
Gartner IT Symposium/Xpo Orlando 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
The leaders chasing short‑term savings keep learning that the quickest road to ROI often detours through damage control. Klarna may be the loudest, but it won’t be the last.
Run the real numbers before firing the humans. If the math survives, proceed. If not, congratulations. You just dodged your next expensive mistake.
Editor's Note: Read more thoughts on how to set your AI pilots up for success:
- How Businesses Are Turning Generative AI Into Measurable Value — Successful GenAI rollouts share at least one thing in common: the initiative starts by identifying and aligning with a core business goal.
- AI Isn't Magic. Prepare Your Data and Your People First — While powerful, AI isn't magical or instantly transformative. But with some prep work, you can deliver value.
- If We Want AI to Help HR, HR Has to Join the Conversation — Engineers are designing AI systems to address problems that are rooted in the very systems HR understands best.
