ChatGPT should make customer service easy. Why is it still so hard?

The strangest thing happened to me recently. I contacted a customer service department and I liked it. I sent an email, received a prompt response, and received a refund. The most remarkable thing about the positive problem-solving experience was the fact that I couldn't tell if anyone other than me was involved.

I realized, albeit briefly, that the prophecies were finally coming true. AI finally made it easier for me to complain to companies and get results. At least that's what I wanted to believe.

Customer service should be one of the things that AI can easily do. In fact, this one good experience was made possible by an AI-first company called Intercom. You have an AI agent called Fin which handles most of its customers' inquiries. Why not everyone?

“I'm confident that a lot of the current customer support that's done over a phone or a computer is going to cause these people to lose their jobs, and that can be done better by an AI,” says Sam Altman, CEO of OpenAI said Tucker CarlsonIn September of all days.

Altman is hardly the only Silicon Valley executive championing customer service automation. Last year, Salesforce Cut 4,000 customer service jobs in favor of AI tools and Verizon started a chatbot powered by Google Gemini as the front door for customer service. Then there is Klarna, their CEO bragged about replacing people with AI before going back last May and launched a recruiting campaign to hire more human customer service representatives.

There's the problem. It turns out that AI, and generative AI in particular, is really good at some things…until it's not. That's why you are Still need to check the facts Everything ChatGPT tells you and why, including chatbots are good at diagnosing certain medical conditions cannot replace human doctors. When it comes to customer service, AI can be good at simple tasks like issuing refunds, but bad at handling more complicated cases, especially when customers are upset and could benefit from some human empathy. Quote Moderator“It works 60 percent of the time, every time.”

Yet human customer service agents are losing their jobs to AI in large numbers, and have been doing so for several years in the United States and abroad. Whether for cost reasons or to seem cool, many companies introduced AI-powered chatbots as the first point of contact for customers, only to find that customers actually hate the concept. Now these organizations are withdrawing from these plans, it is said Brad FagerResearch Director for Customer Service and Support Executives at Gartner.

“The idea that you could replace your workforce is really not feasible and not even preferable,” Fager told me, pointing out that executives may think that replacing human agents with AI is a good way to cut costs. “The reality is that it just doesn’t work.”

There's also evidence that customers simply don't like interacting with AI. This was the result of a Gartner survey from 2024 61 percent of customers would prefer it Companies are not using AI for customer service at all and 53 percent of them would consider switching to a competitor if they did. As Fager explained to me, Gartner's broad view is that AI and automation are important will transform the future of customer service, but that people will play a big role in this change. And to the delight of many customers, much of the AI ​​integration will take place on the backend, helping human agents do their jobs better rather than leading interactions. Customers themselves may never know that AI was involved.

This approach reminded me of a study I read a few years ago by researchers at MIT and Stanford who studied how generative AI increased the productivity of call center workers. This was particularly the case with the less experienced agents. With access to an AI tool that made real-time suggestions for handling calls, agents were able to resolve 14 percent more cases per hour. The tool was trained using data from more experienced agents and could even help novice agents be more empathetic with customers.

Compare this to what you've probably experienced with chatbots: the AI ​​version of a phone tree. Here, you ask a customer service bot for help and are presented with a menu of options that ask you to narrow down your query to get to the right, likely AI-powered, agent. It's a slightly updated version of the annoying phone tree that asks you to say or press one for billing, two for tech support, etc.

These front-end solutions for identifying customers and their needs are essentially AI tools bolted onto old customer service systems, and they are terrible. Werner KunzProfessor of Marketing at the University of Massachusetts Boston, argues that many companies do this just to do something with AI.

“It doesn’t work very well,” he told me. “The failure rate is far too high compared to the older systems, and if companies are currently using AI to do that, I think it is destroying customer relationships.” Kunz added that using AI on the backend would deliver better results in a safer environment and also, “Who cares if you use AI or not?”

This brings me back to my recent, surprisingly positive customer service experience. I contacted Intercom, the company that developed the software, and confirmed that it was an AI agent that solved my problem. There was no analogue to the phone tree and, in some ways, no struggle with a chatbot to reach a human agent. Fin, the AI ​​agent, registered my complaint, offered me a solution in a human-sounding email – even using emojis in the right context – and closed the case before I even thought of getting angry.

It wouldn't be entirely accurate to say that customer service is finally starting to get better thanks to AI. As Kunz and Fager explained, many companies make a mistake by using AI for the wrong purposes or integrating it into legacy systems. However, Intercom co-founder and chief strategy officer Des Traynor says using AI is the best way to give customers what they want: instant results.

“They don’t want to wait,” Traynor said. “It's the same reason people google before picking up the phone: people just want an instant solution to problems, and that's exactly what AI provides.” He added: “It’s just categorically better for users – when it works.”

Traynor admitted that AI ushered in an era of software that had people wondering if it worked, and that problem guided Fin's development. He said his company “spent a phenomenal amount of time developing an AI evaluation engine” and “put every release through its paces” to make sure Fin isn't hallucinating or doing anything wrong. As a result, Fin resolves one million customer queries per week with a resolution rate of 67 percent, which is not 100 percent, but Traynor said that number is increasing by 1 percent each month. He acknowledged that some interactions require human intervention, but in most cases AI can do the job better. In my case that was true.

The big problem is that as a consumer, you don't necessarily get to decide how a particular company handles its customer service. There's also something of an income equality gap between the haves and have-nots, with larger companies like Amazon able to invest more and provide better customer service, while small companies like local utilities simply do their best.

What is clear, however, is that change is taking place. There are signs that it is getting easier to complain to companies, but there are also clear signs that many companies, despite wanting to make it easier, will continue to make it difficult. AI is supposed to help make things work better, but only if it can first stop making things worse.

A version of this story was also published in the User Friendly newsletter. Register here So you don't miss the next one!