No, the header is not just a clever way to drive traffic to this article. It's a fact.
Modern AI, including ChatGPT, is based on neural networks, powered by deep learning. The key technologies that enable the approach were invented years ago. The “transformer” machine learning model was invented by Google already in 2017, and it enables self-supervised learning. The incredible abilities we now see are the result of self-learning AI that independently teaches itself by scanning vast amounts of data, billions of parameters, without being instructed by a human. It learns languages, facts, and generates understanding based on the meaning of concepts.
So what makes ChatGPT, Dall-E, Midjourney or Stable Diffusion a breakthrough at this point in time? Why didn’t it happen earlier?
I believe there are two main reasons. The first is that we now have sufficient processing power at scale. Cloud computing and the massive data centers built by the cloud giants allow for concurrent processing of vast amounts of data, and the data has also been made available on these cloud platforms.
The second is simply because the creators of mentioned services decided to release their service to the public. Several of the upcoming contenders, and you will see loads of these services launched to the public during 2023, have been a bit more cautious - and rightfully so.
The technology is incredibly powerful, but with such power comes great responsibility. We need to make sure that we protect individual integrity, avoid disinformation campaigns, bias, hate speech and avoid causing security concerns.But the rabbit is out of the hat and there’s no turning back. We are yet to see setbacks and how these issues will be solved before the technology is widely applied.
AI will most certainly revolutionize customer care and CX in the immediate future. However, it’s not going to replace the need for humans. If anything, it can easily take care of common tasks with clear business rules.
Let’s reflect on some key capabilities needed for an agent to deliver excellent customer service:
Example: Imagine a customer service agent who deals with invoice inquiries for an online retailer. A service like ChatGPT, that has been trained on massive amounts of general information from the internet, could easily and out of the box deal with the first two steps. It could also express itself in a courteous way (step 3) and give guidance around generic steps associated with interactions between a customer and an online retail company in invoice inquiries (step 4). But this is as far as it currently goes.
So, to make ChatGPT style AI truly empowered to deliver (nearly) all the capabilities of a human agent, we need to do more work:
We are already delivering AI-powered solutions to enhance CX and performance, with many success stories. There are voice and text bots that solve common tasks, do triage of customer queries, and route customers to the best service or agent. Some of our agents are assisted by AI to give live, context specific guidance on where to find more information or even suggest the “next best action”. We have AI monitoring and learning about sales performance and agent behavior, ready to coach new employees on best practices to win more deals or cross/upsell to customers. We have AI screening of job applications that helps us find and recruit the right talent. All of these applications of the AI technology are there to enhance customer experience, lower the cost of operations, and increase agent experience.
Another trending topic is AI-powered, real-time translation of conversations between agents and customers, by chat or voice. This is a potential game changer for our business that’s showing great promise from our first implementations that are live today. This means that, with the help of AI, an Arabic speaking agent from Tunis can support a customer who speaks German. It helps us eliminate language barriers. Even though the two people are speaking different languages, they’re able to communicate in real-time, focusing on the topic and the outcome of the conversation. It’s a solution that will help us solve talent shortages in some markets and significantly lower the cost of service.
At the end of the day, I’d conclude that there are plenty of situations that require human touch. The distinct ability to do what’s right and maybe even bend the rules a little. This capability is still very much human and requires knowing how to tune into the customer's situation and make judgment calls that are still out of reach for AI. Robots are not yet ready to completely replace us humans. Yet.
Stefan joined Transcom in September 2018, as Chief Technology Officer. Stefan's career spans notable positions in well-known brands as well as startups, such as Business Area Manager for Cybercom Group, CIO for Sweden’s largest MSO/ISP Com Hem AB, and most recently, VP Connected Consumer Solutions Electrolux.