I often hear claims that AI chatbots are so good now, Customer Support Representatives will be among the first jobs replaced by AI. If you believe sensational headlines and willrobotstakemyjob.com more than two thirds of CSR jobs are going away in the next two decades. To some extent it’s already happening. Countries like Bangladesh and the Philippines whose economies rely on this type of outsourced contact center labor are scrambling to adapt. But is this change good thing? Is the customer experience (CX) actually improved by AI?
It seems nearly every time I’ve called a customer service number in the last decade or more, I’ve been greeted by an automated phone system designed to push me towards the website or generally get me off the phone as soon as possible. Only if I can successfully navigate the maze of menus and make the right number selections do I get to speak to someone about my issue. So impenetrable are some of these phone trees, I have sometimes had to resort to requesting the all-too-welcoming sales department just so I could ask the poor sales rep to transfer me to a CSR.
Now, mind you, I am a millennial. I grew up with the internet. I know how to use the company’s web portal (though I may still refuse to install their app). So if I’ve run into a problem that forces me to pick up a phone, I have exhausted all other options, and I will need to speak to customer service. Every phone tree I have endured has only wasted my time. I may be going out on a limb here, but I believe most, and a growing number of people fall in this category and feel the same way. Of course, I’m sure there are also plenty of people out there that believe dialing a phone number is fastest way to get an account balance or store hours. Perhaps the automated system is actually useful to them. Bless their souls.
For the latter group, agentic AI will be a God send. It is a far better experience to talk to a conversational chatbot than the phone trees of today. And with Retrieval Augmented Generation (RAG), AIs can access your specific data, balances, payment history, etc. and discuss it with you. No more frustration waiting on long menu readouts so you know what number to press when you already know what you want. No more simply telling you to go to the website and hanging up. In fact, with such improved CX I imagine we could see more people opt to call in to an agentic AI interface to complete mundane tasks over using the web app.
However for people like me who only call in for real issues, I reserve some concerns. The AI CSR chatbots I’ve had to deal with on websites today fare little better than automated phone systems in actually being useful. Granted, these simple purpose-built chatbots that are trained on the website’s FAQ section and apparently little else don’t really compare to the possibilities that agentic AI brings to the table.
When my internet went down recently, not only did I know how to restart and read lights on the router and ONT - that much is on my ISP’s website, I also knew how to test cables and ping network devices. So when I called my ISP I was ready to confidently tell them the issue was on their end. You may already be getting flashbacks to your time in the trenches with ISP customer service. If so, you understand the struggle I went through to actually get a tech on the line, who of course asked me to walk through the basic troubleshooting steps posted on their website once again.
Could AI have made this interaction any better? Theoretically, yes. First of all, AI is not going to be able to solve the physical connection issue at the ISP’s fiber distribution hub. For that they are still going to need to send a tech out, a human still has to be involved. But theoretically, an agentic AI with RAG could look at my account, see my connection is down, confirm I’ve done the at-home troubleshooting steps, quickly set up a ticket, and schedule a tech to come out without a technician ever having to come to the phone. But in practice? We have to set the right prompts and goals for AI agents. If our customer service chatbots like the automated phone systems and AI chats of today start with the assumtion that the customer is probably looking for simple information that is already on the website, people like me are going to have a terrible experience, and an even harder time getting proper help when the chatbot becomes the only point of contact.
Current general chat interfaces like Claude, ChatGPT, and Gemini are actually quite helpful, because they are designed to be generally knowledgable, very agreeable (sometimes to the point of sycophancy), with no ulterior motive to end a conversation. They are also quick to admit when they can’t do something. If these qualities are adopted by agentic AI filling customer service roles, they should do quite well. But that last part is the catch. If a chatbot has to admit they can’t do something, then we still need humans to do it. If we are agreeable to this, that means our call centers aren’t completely going away any time soon.
So will AI take over customer service? Not completely. AI will certainly revolutionize customer service, but I think many companies will realize that a hybrid approach is superior. AI can summarize conversations, perform sentiment analysis, and categorize calls in ways that can accurately route calls, improve response times, and provide data that feeds back into the system to make it even more effective over time. These concepts are emerging in the field of Customer Experience (CX) and I believe they’ll gain significant traction. Sure, some companies will likely be overconfident in their AI investment and try to ditch the contact center entirely, to their, and their customer’s detriment. But, over time, the virtues of the hybrid AI augmented CX approach will win out.