OMQ Is Automating Inefficiency Out of Customer Service

Customer service can be tricky and tedious—both for those trying to get answers and those trying to provide them. More often than not, a person who calls or emails a business for help with its product or service has a question that has been answered many, many times before. And yet, there can be a great deal of inefficiency standing between the customer and what will turn out to be a simple, preexisting answer: long wait times, badly designed websites, and the endless variation in ways that people ask about and describe identical problems.

Matthias Meisdrock became acutely aware of these problems at his first job where, he says, he “heard the company’s service employees in the office next door answer the same customer questions with the same answers over and over again.” That experience “gave rise to the idea and goal of developing a software that would automate these service processes.”

In 2010, Meisdrock and Sven Engelmann founded OMQ GmbH with the goal of creating artificial intelligence software capable of understanding and responding to customer inquiries in real time. The Berlin-based startup developed its technology in cooperation with several partners, including the German Research Center for Artificial Intelligence and the Universität Heidelberg, with whom they worked on AI projects focused on natural language understanding and deep learning.

No matter what industry a company is in, OMQ’s software is designed to assist with customer service through multiple communication channels. Describing the product OMQ Contact, Meisdrock says, “As soon as a customer enters a question in the input field, suitable answers appear in real time next to the form. In this way, the customer’s question is automatically answered before it is sent to customer service.”, providing fast support and frequently avoiding the need to generate a customer service ticket. Similarly, OMQ Reply provides a self-answering email system that recognizes recurrent request and answers them automatically. The answers are entered in a ticket and marked as closed.

As the AI fields more requests, it gets better at recognizing when vaguely or confusingly phrased inquiries are in fact common questions to which it has an answer. Meanwhile, OMQ Help functions as “a dynamic FAQ,” moving answers to the top of the page according to what questions customers are currently asking most often (say, during a seasonal promotion), and guiding visitors to solutions via an intelligent autocomplete navigation system.
The OMQ Chatbot automatically asks queries if there are several possible answers or connects the customer with a service agent if the question is too individual. The holistic OMQ System uses the same knowledge for all communication channels and all products learn from the requests and feedback of the other products.

OMQ is still actively following the latest AI research and looking for innovations to integrate into its products. Most recently, it began implementing an attention-based transformer model, which is pre-trained on general text data and then fine-tuned for customer service data, allowing it to classify new, previously unseen requests without explicit training. Meisdrock says that, as a lean startup, OMQ has “worked without the help of investors since its inception” and “takes great care to quickly introduce and establish the software on the market,” gaining and maintaining customers through testing and iteration. With over 100 clients and 930,000 processed requests per month, Meisdrock says that OMQ is moving toward its vision of “answering every kind of service request only once, or [in other words] automating 100 percent of customer service.”