A TikTok video opened my eyes about CX in the age of AI.
Agentic AI,
Conversational AI,
AI translation,
airlines,
agent training,
AI,
GenAI,
Generative AI,
Retail,
CX,
Chatbot,
Published on Tue Jun 17 2025
Updated on Fri Aug 08 2025
4 minute read
A few days ago, I came across a TikTok video that made me smile. But only up to a point. A young woman, accompanied by her grandmother, pulls up to a KFC drive-thru in Australia and tries to order a simple sandwich with barbecue sauce. On the other end, however, there isn’t a human being. There’s a voice assistant powered by artificial intelligence, recently introduced to “streamline” the ordering process. Instead, the AI repeatedly misunderstands and adds multiple barbecue sauces to the order instead of the sandwich. KFC clarified that this is a pilot phase and that, if there are issues, it’s always possible to speak with a real staff member. But the concerns remain. Incidents like this are worrying warning signs for anyone blindly relying on automated customer service. The illusion of drastically reduced costs, coupled with the pursuit of higher productivity, has made many forget the first rule of using generative AI: never fully trust the response from a chatbot.
The Klarna story.
Sebastian Siemiatkowski, CEO of fintech company Klarna, learned this the hard way. “We stopped hiring new staff a year ago,” Siemiatkowski told Bloomberg last December. Just a few days ago, in an interview with CNBC, he added that Klarna’s workforce has shrunk by 40% since 2023 - in his words, “naturally.” How? By replacing departing employees with AI. The result initially seemed spectacular: Klarna proudly announced that its customer service chatbot could now do the work of 700 human agents - and do it better. Siemiatkowski even used an AI-generated clone of himself to present the company’s Q3 2024 financial results. I’m not joking: the announcement was made by a digital “Sebastian.” Only - surprise - things didn’t quite work out as hoped. “Unfortunately, it seems cost was the main factor driving the strategy,” Siemiatkowski admitted to Bloomberg a few weeks ago. “In the end,” he confessed, “we ended up with lower quality.” As a result, after an aggressive push for customer service automation, Klarna has backtracked: it will start hiring human agents again, this time as freelancers working remotely. Let’s hope the Swedish company doesn’t repeat the mistakes it made recently in Riga, Latvia, where it outsourced customer service to a major Indian company, Tech Mahindra. An investigation by Swedish newspaper Aftonbladet (you can find it
Offering a “human exit strategy”.
Creating a “human exit route” from the AI maze is crucial. No one wants to feel trapped in a panic room full of bots, especially during stressful moments or when problems get complicated. To truly thrive in a world where customer experience is increasingly intertwined with AI, companies may need to take a step back, not to halt progress (which won’t stop anyway), but to observe it more clearly. And perhaps to keep in mind three simple truths:
- Empathy can’t be programmed.
- Algorithms help, but they don’t truly listen.
- AI, on its own, is not enough.
Concrete examples.
A concrete example of this is “Ask,” an AI assistant designed to help AppleCare agents (the ones answering customer questions and solving problems every day). The system is simple yet smart: agents input (or rephrase) a customer question, and the AI searches Apple’s internal knowledge base to generate a relevant, up-to-date, and helpful answer that can be shared in a chat or phone call. In practice, “Ask” cuts down the time agents would spend browsing documentation, searching databases, or asking a senior colleague. Not only that: Apple encourages its technicians to use this tool as a first step when faced with a complex or unfamiliar problem. Similarly, Comcast (one of the largest cable and internet providers in the U.S.) has introduced a system called Ask Me Anything (AMA), designed to assist its customer service reps. This chatbot allows agents to ask questions to a large language model during live customer conversations. Instead of replacing the human, AMA acts as a silent assistant, suggesting quick and relevant responses and greatly reducing the time needed to find technical or procedural information. The integration of this assistant led to a 10% reduction in average handling time for calls that required research and even improved the perceived quality of the customer experience. The strategies of Apple and Comcast show us that when human empathy and critical thinking are combined with the speed and precision of intelligent systems, customer experience becomes smoother and more satisfying. It’s not about choosing between humans and machines. It’s about making them work together
Read more
Created at Wed Apr 29 2026
4 min read
Stop automating pain: why AI won't fix your customer service issues.
We make unconscious choices several times every single day. Most people rarely stop to think about them because they are unconscious - it requires focused effort to stop and think precisely about what you are doing. Driving is a good example. When you first learn to drive a car, you need to think about each action, but it eventually becomes natural and fluid.
Earlier this week, I was thinking about this in the context of customer service teams. It’s one of those subjects that works well at the
