Since Generative AI became widely available, the customer experience industry went from skepticism to infatuation to disappointment. Some feared unsupervised AI could ruin companies, others hoped AI could replace costly staff. The reality turned out to be much more nuanced for all. Neither the worst fears, nor the promises of total automation came true.
As a hands-on leader who started as a customer support consultant 15 years ago, I've watched these waves come and go. I built this company with people, for people so I wanted to know what this reality check actually tells us about the real opportunities for AI in CX.
I didn’t have to guess: we started exploring AI in CX early on, around 2020. That gave us our own cases, our own data, and a unique take on what was going on. The TL;DR conclusion was: the AI was neither a friend nor a foe. AI is a powerful tool with a transformative potential when people administer it right.
Our experience showed the limitations of AI in customer service and proved that technology delivers real value when humans stay in charge. On the other hand, when leaders let common misconceptions drive their AI integration, businesses end up paying twice.
The Illusion of Flawless Automation
When conversational AI for customer service emerged, business leaders hoped it could be more efficient than their staff. They viewed AI as a modern version of an assembly line conveyor belt: fast, flawless, never tired. Until the line started pumping out broken parts at scale.
In 2026, the problem of AI making mistakes and malfunctioning became more than a niche discussion. When BBC and EBU journalists analyzed responses from popular chatbots, they revealed almost half of them had accuracy issues. While some wrong answers may seem like a minor annoyance, others carry real operational risks.
OECD monitors media reports on AI incidents and hazards. According to the findings, the monthly numbers surged tenfold, jumping from around 50 in early 2020 to nearly 500 by January 2026.
From some of those media stories, we know that AI can hallucinate, swear, and give dangerous medical advice. So when our clients ask for AI to scale, we are transparent about the limitations of this technology and offer to combine it with human expertise.
Within our human-in-the-loop model, AI uses knowledge bases, analyzes customer sentiment, predicts their behavior, and drafts responses that a human can approve and send. A hybrid model offers a different type of collaboration: AI autonomously resolves routine queries without consulting the specialist, only routing complex, sensitive cases to humans. In both cases, customer support gets the best of the two worlds.
Compliant hybrid support means AI and human agents work within a system built to protect customer data at every handoff. Whether AI drafts a reply, resolves a routine ticket, or hands a case to a person, each step follows recognized data protection and security standards. Human oversight isn't just there for quality: it's the checkpoint that keeps sensitive data and legal risk in the right hands.
Cutting Headcount Creates Bottlenecks
After the business world hesitantly accepted that AI needed human supervision, leaders still hoped hybrid customer support would mean hiring fewer people. Some took this idea to extremes and traded quick savings for long-term damage.
When a Swedish firm, Klarna, went all-in on AI chatbots to shrink headcount and handle millions of queries, reality caught up quickly. Prioritizing cost-cutting over human empathy tanked service quality, and the company ended up re-hiring people. The AI couldn’t replace human touch, and Klarna paid for it twice.
What if businesses used generative AI customer service to amplify specialists instead of replacing them? Here’s an example from our practice. When an email validation platform, Mailcheck, wanted to prepare for a spike in support requests, they reached out to us. We paired our human expert with an AI agent, so that AI handled 81% of routine queries independently and handed complex cases to a human.
This strategy proved worthy, and two months in, we added two more teammates to keep up with the new volume. AI absorbed the ticket volume that came with Mailcheck’s growth, while human experts kept the quality consistent.
As a business leader, I totally understand the urge to cut costs. But real cases prove that actual savings come from empowering teams with AI, not from layoffs.
The Low Costs Delusion
The more business leaders tried to save costs by replacing humans with AI, the more cases emerged proving AI wasn’t automatically cheaper and better. In some situations, it was either cheaper, or better, or neither.
Look at this noble intention of cutting administrative costs in the public sector. When New York City launched the MyCity AI chatbot, they wanted to provide more accessible assistance to small businesses. But the bot ended up routinely advising business owners to break the law.
Budget-driven AI without human oversight introduces severe legal and financial risks. Automated mistakes land companies in court and wipe out brand’s trust overnight. To avoid that, we offer a model where AI optimizes costs by taking on routine tasks, and transfers tickets to people as soon as it identifies a complex problem.
When a macOS software developer, Cocoatech, needed to tackle a massive support backlog, replacing their team with standalone AI wasn't an option. Their technical edge cases required real expertise.
So we set up AI to take routine queries while handing off complex technical tasks to human specialists. We managed to automate 81% of incoming chats, drop resolution times, and increase their customer satisfaction. Since Cocoatech didn’t need to hire more people, they got the savings they needed.
AI can indeed be cheaper, and it can certainly be better. But left unchecked, it can easily turn worse and far more expensive, losing the unnecessary battle between AI vs human customer service.
At SupportYourApp, compliance sits inside the hybrid model itself. We hold PCI DSS Level 1 certification, meet ISO/IEC 27001 standards, and follow GDPR, CCPA, and HIPAA requirements across both our AI and human workflows. AI handles routine, lower-risk tasks inside these boundaries, while trained specialists take over sensitive or complex cases, so client and customer data stay protected no matter who, or what, is handling the ticket.
The Indifferent Customer Trope
With humans overseeing it for accuracy, AI could finally take the messy frontlines. Unlike its supervisors, AI doesn’t burn out, doesn’t take holidays, and operates completely free of cognitive bias. And that was the next illusion the business world went through.
By 2026, the tech world collected a reasonable amount of data proving it was wrong. A popular survey platform, SurveyMonkey, found out that on the American market, 79% of the customers strongly prefer interacting with a human over an AI agent, with an even larger share (89%) supporting a belief that companies should offer the option to speak to a human.
A global advisory firm, Gartner, reached similar conclusions after surveying more than three thousand B2B and B2C customers globally. They found out that 87% of them expected the option to reach a human when using AI customer support.
We also see that in our daily work. Our clients naturally look to AI to scale without expanding teams, but make sure humans stay right where empathy matters most.
For instance, we built a 24/7 multilingual AI voice support for a U.S. law firm assisting migrants. For clients navigating urgent, life-changing status issues across language barriers, that immediate availability became a lifeline.
But we knew that wasn’t enough. Complex legal cases, however urgent, require human judgment and precision. So we set up a system that routes such inquiries from an AI voice agent to dedicated human specialists who go beyond scripts and respond with genuine empathy.
Customers do care who’s answering them. They appreciate speed, but when it matters most, they want to be heard by someone who actually understands. Hybrid customer support is exactly what allows it.
The Dream of Total Replacement
As business leaders, we check AI for accuracy, bias, cost, and empathy for one simple reason: we're asking why AI couldn't replace humans in customer experience. As a CEO in a fast, relentless market, I’m naturally pushing for efficiency. But as someone who’s built the company with people and for people, I know technology is not capable of replacing people.
Tech scales velocity, but only human judgment safeguards value. When businesses deploy unsupervised AI to slash headcount, they trade efficiency for legal liability and broken trust. By 2026, we’ve seen enough cases: analysts collected impressive data proving that real cost-efficiency happens when AI empowers people, not when it replaces them.
Across hundreds of client setups, our practice at SupportYourApp proved the same truth: standalone AI creates expensive chaos. But human-in-the-loop and hybrid models alike offer an effective combination of AI with human expertise that delivers real, long-term ROI.
In my experience, to reach true efficiency in the age of AI, leaders should abandon the myth of complete replacement. Unsupervised AI tools trigger costly errors and damage trust, while people alone can’t keep up with the modern volume. Installing a smart hybrid model fit specifically for your company’s size, team composition, and goals is the ultimate way to win with AI.
