Many people will recognize this: you call customer service and end up in an endless menu of options. “People want to be helped immediately, instead of navigating through a maze of choices,” says Tias. The same frustrations apply to the current generation of chatbots. “These bots often work based on pre-set flows,” Tias explains. “If your question doesn’t quite fit within those options, you won’t get a relevant answer.”
AI and LLMs (Large Language Models) are transforming the way companies communicate with their customers. According to Tias Hazebroek, this technology offers many opportunities for more efficient customer service. It goes far beyond those frustrating chatbots and IVR systems.
The maze of IVR systems
LLMs go further
With LLMs, it’s different. “Instead of being limited to pre-determined answers, LLMs can engage in open dialogue. They understand what you’re saying and respond accordingly, without having to follow a rigid structure. This makes the interaction feel more natural and faster,” says Tias.
A Large Language Model (LLM) is an AI system that understands and generates text. It is designed to process human language and provide natural responses. This allows companies to quickly and automatically answer customer questions without human intervention. By combining this with speech-to-text and text-to-speech technologies, you can even have voice conversations with an LLM.
Multiple languages
“LLMs can handle multiple languages. Whether you speak Dutch, English, Spanish, or Chinese, AI can adapt to all of them.” This enables companies to serve customers worldwide. “Additionally, a bot doesn’t get sick and is always available. You can scale your business endlessly and ensure 24/7 availability. Waiting queues will also become a thing of the past,” says Tias.
Speed and efficiency in customer contact
AI LLMs can be set up to handle common customer queries more quickly, even for questions that don’t follow a standard script. An LLM can autonomously and efficiently resolve open-ended questions like, “When do you still have two balcony tickets available for Herman van Veen’s show?”
This allows employees to focus on more complex issues that require human intervention. “Customers expect immediate answers,” says Tias. “With the help of our AI solutions, you can meet these expectations. This increases efficiency and reduces the pressure on customer service departments. It’s a win-win: it improves customer satisfaction and often leads to cost reductions as well.”
"Shit in, is shit out"
To achieve this, the quality of the responses is a key focus. AI is only as good as the data you feed and train it with. “If you train an LLM with incomplete or incorrect data, you’ll get inaccurate answers,” explains Tias. “Garbage in, garbage out. Continuous monitoring and improvement are therefore essential.”
The balance between AI and human interaction
A hybrid approach is a good starting point. “In the first phase, you use an LLM for simple and routine questions. At the same time, you ensure that a human agent is available for situations that require more empathy and nuance or when customers prefer not to interact with an LLM. That option must always exist.”
Privacy
One aspect to consider is the privacy of customer data. LLMs learn from the data they receive, which raises questions about how this data is collected, used, and protected. “Customers need to trust that their data is stored securely and only used with their consent,” says Tias. Companies must therefore focus not only on the efficiency and business benefits of AI and LLMs but also on the ethical and legal aspects. Sound of Data is happy to assist you with this.
This interview is a shortened version of an article from Twinkle (08-2024).