Klarna's AI Reality Check Shows What Happens When You Move Too Fast
PLUS: Smart AI deployment readiness + Quick customer journey mapping
Start every workday smarter. Spot AI opportunities faster. Become the go-to person on your team for what’s next.
🗓️ September 11, 2025 ⏱️ Read Time: ~5 minutes
👋 Welcome
Here's a story that should make every CX leader pause: Klarna, the poster child for AI in customer service, just admitted they went too far too fast with AI and are now course-correcting. Meanwhile, Albertsons is proving that conversational AI works when done thoughtfully, Adobe is making it possible for any business to build AI agents without coding, and Lennox is putting AI directly in the hands of their technicians. What ties these stories together? The companies succeeding are the ones moving deliberately, not desperately.
📡 Signal in the Noise
Every story today points to the same lesson: AI deployment isn't a race to replace humans as quickly as possible. It's about finding the right balance between automation and human connection that actually improves customer experience.
🎯 Executive Lens
Here's what I'm seeing: the companies getting sustainable results from AI aren't the ones cutting the most jobs or automating the most processes. They're the ones who understand that good AI deployment takes time, testing, and constant adjustment based on what customers actually want.
Stories That Matter
🔄 Klarna admits it went too far with AI and is now hiring people back
So here's a reality check that every CX leader needs to hear. Klarna's CEO just admitted they "over indexed" on using AI to cut costs and are now course-correcting. The company that made headlines for replacing 700 customer service jobs with AI is back to hiring people. Why? Because while their AI chatbot can handle simple queries in two minutes instead of 11, complex issues like identity theft still completely stump the technology. Investors aren't cheering the cost savings—they want growth and better customer service.
Why this matters: When the poster child for AI customer service admits they moved too fast, it's a wake-up call for everyone. Cost cutting isn't a strategy; solving customer problems better is. The companies that figure out the right balance of AI and human expertise first will have the real competitive advantage.
Try this: Before automating any customer interaction, map out what happens when the AI fails. Do customers get stuck in loops, or can they easily reach a human who actually understands their problem? That failure scenario is just as important as the success scenario.
Source: Reuters
🛒 Albertsons proves grocery shoppers want to talk to AI, not click through menus
Google just launched something called Conversational Commerce agent, and Albertsons is the first big retailer to actually use it. Here's what's interesting—instead of making customers click through categories like "produce" then "fruits" then "apples," people can just ask "What apples are good for baking?" The AI knows Albertsons' entire product catalog and can have back-and-forth conversations to help customers find exactly what they need. Early results show that over 85% of successful shopping sessions start with open-ended questions, and customers often add extra items they wouldn't have found otherwise.
Why this matters: Think about how frustrated your customers get navigating your website or app. If grocery shopping—one of the most routine activities people do—works better with conversational AI, what does that tell you about how customers want to interact with your business?
Try this: Next time you're on your own company's website, try to find something specific using only the search function. Count how many clicks and menu selections it takes. Then imagine if customers could just ask for what they want in plain English and get it immediately.
Source: PR Newswire
🎨 Adobe makes it possible for any business to build their own AI agents without coding
Adobe just made their Agent Orchestrator generally available, along with six different AI agents that businesses can customize for their specific needs. Here's the cool part—you don't need developers to set this up. The system includes agents for building customer audiences, optimizing journeys, running experiments, and handling support issues. Coming next month, they're launching something called Agent Composer that lets you adapt these agents to your brand voice and business goals just by configuring settings.
Why this matters: Until now, building custom AI for customer experience required months of development work and technical expertise. Adobe's betting that the future belongs to businesses that can quickly adapt AI agents to their specific customer needs without waiting for IT projects.
Try this: Think about the customer experience tasks your team does manually that follow predictable patterns—building audience segments, analyzing why customers drop off, or handling routine support questions. These are exactly what Adobe's agents are designed to automate.
Source: CIO
🔧 Lennox puts AI agents directly in technicians' hands, saving time in the field
Lennox launched AI agents for both their HVAC technicians and homeowners, and the results are pretty impressive. More than 7,000 technicians have registered for the tool, which helps them troubleshoot equipment, look up warranty info, and find repair parts by just asking questions. The AI recognizes over 250 error codes and can guide technicians through repairs step-by-step. For homeowners, there's a separate agent that helps with basic troubleshooting and finding local dealers. Both tools work in English, Spanish, and French.
Why this matters: This shows how smart companies are putting AI where the work actually happens—in the field with technicians and at home with customers. Instead of making people call support or search through manuals, the AI brings the answers directly to where they're needed.
Try this: Map out where your customers and employees actually need information when they're trying to solve problems. Are they at their desk with a computer, or are they standing in front of a broken machine with just their phone? That's where AI assistance needs to be.
Source: PR Newswire
📊 MIT study reveals what actually works when integrating AI into customer experience
MIT Technology Review just published research on how companies are successfully balancing AI with human customer service. The key finding? The most effective organizations treat AI as a collaborative tool that enhances rather than replaces human connection. They're also discovering that while customers are warming up to AI, excessive personalization can make people uncomfortable, and overly empathetic bots feel fake. The companies getting this right are being transparent about AI use and giving customers clear ways to reach humans when needed.
Why this matters: This research cuts through the hype and shows what actually works in real customer service deployments. It's not about replacing humans or building the most advanced AI—it's about finding the right balance that customers actually prefer.
Try this: Before rolling out any customer-facing AI, test it with real customers and ask them directly about their comfort level. Find out where they want AI assistance and where they absolutely want a human. The answers might surprise you.
Source: MIT Technology Review
💰 Quick hits:
• Replit raises $250M at $3B valuation as AI coding tools prove their worth for developers building customer-facing applications
• Perplexity secures $200M at $20B valuation, showing massive investor appetite for AI search that provides direct answers instead of link lists
• Born raises $15M to build social AI companions that bring people together instead of isolating them with one-on-one chatbot relationships
🎯 Prompt of the Day
Title: AI deployment readiness assessment for customer-facing features
You're helping me evaluate our readiness to deploy conversational AI in customer interactions. For [YOUR SPECIFIC CUSTOMER TOUCHPOINT], analyze and create an action plan covering:
1. **Natural Language Needs:** What questions do customers ask most often, and how do they actually phrase them in real conversations?
2. **Context Requirements:** What information does AI need to access to give helpful answers (product catalogs, customer history, inventory, etc.)?
3. **Conversation Flow:** Where should conversations start with AI, when should they escalate to humans, and how do we make those handoffs smooth?
4. **Success Measurement:** How will we track whether customers prefer conversational AI over current options like search, menus, or phone calls?
5. **Implementation Plan:** Start with the simplest, highest-impact conversations and build complexity gradually.
Focus on making interactions feel more natural and less like work for customers.
This helps you design AI conversations that customers will actually want to use instead of avoiding. Use it to create deployment plans that start with genuine convenience and build from there.
⚡ Try This Prompt
Map our customer journey for AI conversation opportunities. Identify: 1) Where customers currently struggle to find information quickly, 2) Which interactions involve repetitive back-and-forth that could be streamlined, 3) What happens when customers can't find what they need through self-service, and 4) Which customer questions follow predictable patterns. Rank opportunities by customer frustration level and ease of AI implementation.
This reveals where conversational AI can remove the most friction from your customer experience while being realistic about what you can actually deploy successfully.
💭 CX Note to Self
The best AI deployments aren't the fastest ones—they're the ones that customers actually prefer over what came before.
👋 See You Tomorrow
Klarna's admission that they moved too fast with AI should be required reading for every CX leader. It's a perfect example of why the companies that take time to understand what customers actually want from AI—not just what the technology can do—will build sustainable competitive advantages.
What's one place in your customer experience where you could test AI assistance without risking customer relationships if it doesn't work perfectly? That's probably your best starting point.
—Mark
💡 P.S. Want more prompts? Grab the FREE 32 Power Prompts That Will Change Your CX Strategy – Forever to start transforming your team, now. 👉 FREE 32 Power Prompts