From Cuts to Comebacks: Look at How Companies Are Rethinking AI in Customer Support
AI has long been hailed as a transformative force in customer service. After all, it helps boost efficiency, streamlining workflows, and handling scale that humans simply can’t.
However, at some point, companies must ask this question.
Does automation come at too high a human cost, and how does it affect customer trust?
Two compelling case studies emerged recently.
Case Study 1
Salesforce CEO Marc Benioff stated that AI bots have taken over approximately 4,000 customer service roles, reducing the number of support staff from 9,000 to 5,000.
Customer interactions are now managed equally by AI and humans, with no decline in customer satisfaction (CSAT) scores.
Salesforce says automation didn’t just make them faster; it also helped them finally dust off and work on sales leads that have been sitting untouched for years.
Case Study 2
Klarna, the Swedish fintech, initially cut staff and leaned heavily into AI, claiming its chatbot could handle the workload of 700 full-time agents.
But by the middle of 2025, the company had changed its mind and moved employees back to customer support.
Klarna’s CEO acknowledged the mistake, stating that an overemphasis on cost reduction had compromised quality and empathy, ultimately leading to a shift back to a human-centered approach.
These two situations illustrate distinctly different perspectives on what AI is doing in customer service today.
Salesforce’s AI-Based Approach
The main goal of Salesforce’s approach is radical automation.
Agentforce, their AI platform, currently handles half of all support conversations.
CEO Benioff referred to it as an agentic system, an AI that can break down and execute tasks, increasing productivity and even covering a backlog of 100 million unaddressed leads.
He describes AI and humans working in tandem, where humans step in only for complex issues under an omnichannel supervisor model.
What is the payoff here?
Absolutely no drop in customer satisfaction, lower costs, faster, and scalable support. Analysts suggest this may become the model for future enterprises.
Yet, there’s an undercurrent of concern.
What happens to human opportunity, skill development, and the emotional touch with customers?
Klarna’s Course Correction
Klarna’s experiment with automation was equally bold, but the outcome was different. Customers perceived AI responses as robotic, inflexible, and frustrating, resulting in a significant backlash.
Their CEO admitted the mistake by saying that AI can’t understand feelings or deal with complicated problems.
To fix the trust and quality of the customer experience, Klarna transferred personnel back into support roles.
What We Can Learn from the Two Models
| Aspect | Salesforce’s AI-First Model | Klarna’s Human-Centered Pivot |
| Staffing approach | Significant cuts in support staff; automation-heavy | Reassigning staff to support, to strengthen empathy |
| Customer satisfaction | Maintained via efficient automation | Dropped, and later recovered by reintroducing humans |
| Human role | Oversight, escalation, and supervision | Core to quality, trust, and resolution. |
| Strategic focus | Efficiency, scale, handling backlog | Quality, empathy, brand trust |
| Risk | Undermining customer trust, ignoring complex queries | Higher cost, but restored service quality and trust |
Why Is the Balance Between AI and Humans Important?
You might conceive of this as a game with two sides: AI wins or humans win.
However, in reality, the best customer experiences are created when balancing AI and human empathy becomes the focus
Let’s break down why that balance is so critical.
Efficiency vs. Empathy
AI shines at handling repetitive, high-volume queries.
Do you need to reset a password?
Do you want to track a delivery?
Do you want to check your balance?
AI bots can do this in seconds, 24/7, and without the risk of human error.
However, empathy isn’t simply a great thing to have; it’s often what makes the difference between keeping and losing a customer.
Think about how you feel after seeing a fake charge on your credit card. An AI bot may be able to block the card right away, but would it be able to tell that you are stressed out or that your money is safe?
I am sorry, but no.
That’s when a human agent comes in to not only fix the problem but also calm the nerves.
If companies fail to strike this balance, they may create an atmosphere that feels cold and transactional.
Customer Trust Is Fragile
Klarna learned this the hard way. Their AI chatbot was technically efficient. It could manage a lot of questions, but customers thought it was robotic and not helpful when things became hard.
Customers lost trust, complaints mounted, and eventually, Klarna had to move people back into customer care.
By contrast, airlines often get this right. Many use AI to handle simple updates like:
Is my flight on time?
Can I change my seat?
However, when a passenger is stuck overnight because of a cancellation, they are swiftly put in touch with a real person who can rebook flights, find lodgings, and show real care. That combination of speed and sensitivity helps maintain customers’ trust, even when things are challenging.
AI as a Partner, Not a Substitute
Instead of seeing AI as a rival, think of it as a clever helper. It can provide agents with information about a customer’s past, advise on the best next steps, or even generate responses that the agent can modify.
Some banks, for instance, utilize AI to summarize a customer’s previous five transactions and the likely reason for their contact. This way, the human agent doesn’t have to spend the first two minutes looking for the information.
This means that problems get solved faster, the consumer is less frustrated, and the agent is more sure of themselves.
This is commonly called “augmented intelligence,” which means that AI does the hard work behind the scenes while humans make the final decisions.
The Employee Experience Matters Too
It’s not only about customers.
When AI handles the tedious work, agents also benefit.
They can do more important things instead of answering the same “What’s my balance?” inquiry 200 times a day. For example, they may help a small company owner set up a new account or show an older customer how to use digital banking.
That change makes people feel better, lowers the risk of burnout, and makes the support function more strategic.
And when workers are happy, customers are usually happy too.
What is the key takeaway?
AI should never save money at the cost of customer confidence.
The key is to establish an environment where AI takes care of the mundane tasks, humans take care of the genuine relationships, and the two work together to make experiences that are both efficient and caring.
The stories about Salesforce and Klarna reveal a crucial truth.
AI should help people, not take their place.
Automation makes things more efficient and bigger, but people still need empathy, judgment, and adaptability.
This is what a customer assistance model that will last into the future looks like:
- AI handles frequently asked questions and routine tasks.
- People take over when things get sensitive, complicated, or escalated.
- Employees transition from routine to strategic and from transactional to relational roles.
Companies can give their customers efficient, caring, and scalable experiences by combining the best of AI and human strengths.