© scyther5/Getty Images

Chat system Assistify

Smart helper’s assistant in customer dialogue

11/2017 – Assistify helps support staff answer customer questions. But thanks to AI technology, the possibilities go even further. True to the motto: use knowledge in a simple way.

Good communication is crucial – particularly when it comes to applicants, who often have questions that block the application process, such as to whom the application should be submitted. Without an answer, the application would be put on hold and presumably no longer pursued. A live chat enables real-time communication with the support team of DB Dialog. For the colleagues providing the answers, it’s sometimes not easy to find the right answers to complex queries in a short space of time. Even if intelligent knowledge databases with intuitive user interfaces are available to support staff, it doesn’t necessarily mean that standardised answers can be given. This is where Assistify comes into play.

Jan Rudolph, Rüdiger Kurz and Oliver Jägle from DB Systel have already developed the Reisebuddy (“travel buddy”) in collaboration with DB Dialog. While doing this, they also identified the need to establish chat communication in the Helpdesk and combine this with artificial intelligence (AI) that follows the conversation and makes suggestions. It was during the fast-track programme that the idea of the intelligent chat system, Assistify, came about. The self-defined goal was clear from the outset: on the one hand, Assistify should offer relief to support staff in daily work by automatically helping them with recurring tasks and answering queries. On the other hand, this should help resolve customers’ issues more quickly, thereby increasing their level of satisfaction.

Assistify assists the discussion participants in an intuitive chat by implicitly searching in internal DB knowledge sources. While the customer types a question into the chat window, Assistify semantically analyses the conversation in the background and proposes a selection of potential answers to the problem in real time. Even beyond customer support, the tool can offer added value, for example, when experts communicate within a team.

Find answers instead of searching for them

The principle is a little like Google’s auto-propagation function: as you type a term into the search engine, the most frequent search requests containing this string appear automatically – and with fast results, because the search area of Assistify is limited to specific integrated databases, wikis, searches and FAQs. “We integrate special knowledge sources and search for information only within these sources. At the same time, this reduces the maximum search area and returns more relevant results”, explains Jan Rudolph. Experts automatically see a list of answers proposed and can use these as a basis for formulating their own answers. Another goal is also clear: since the tool can remember the answers found, Assistify is to be continuously improved based on which proposals were used, rejected or frequently used. This brings significant benefits to support staff, including external benefits, as it allows even complex requests to be answered consistently in the future by different experts.

“We wanted to develop a scalable product that can be used in a range of contexts”, says Jan Rudolph. This allows Assistify to be used in other fields, too, without much modification to fit different purposes. The aim is always to show suitable additional information in a conversation based on existing sources of knowledge and past conversations. The motto is: use knowledge in a simple way: Thus Assistify learns by saving every conversation that takes place in the system – and uses this knowledge systematically. “This will enable us to link similar questions to these past conversations. This is independent of context and can be applied anywhere”, says Jan Rudolph. One thing was paramount during development: the product should not just have one DB-specific use, but offer a broader basis. The first step, however, is to use AI to extract the important keywords from ongoing communication containing unstructured data, and then perform a search.

Live chat pilot project on the careers portal

A live chat supported by Assistify was launched in August on DB’s careers website. The channel is managed by DB Dialog, whose employees communicate with website visitors with the aim of recruiting talent. Potential applicants can clarify any application-related questions here. The expected goal of answering the questions quickly without interrupting the application process could be quickly achieved.

A possible future scenario would be to use the system to involve the relevant recruiter directly in the communication to help the applicant even more precisely with the matter. In a subsequent step, the collected conversions could be used to develop an even better service, for example, bots could answer frequently-asked questions automatically. Assistify will be enhanced with further functions in the future – even beyond integration in the careers portal. For instance, the knowledge management system of DB Dialog will be connected to propose potentially matching text blocks automatically.

It’s already clear now what significant deployment potential this has within the DB Group: from supporting customer and user support, service desk staff, through to communication within project teams and introducing new systems, new software versions and releases. As of now, Assistify is available within the Deutsche Bahn Group, providing an intelligent communications solution for internal collaboration and for direct customer support. More evidence that digitalisation at Deutsche Bahn has really started to gain momentum.