Contact us
    Digital workspace Microsoft 365 Intranet 03.05.2021

    Machine-learning chatbot helps staff to find information more efficiently

    Advania Finland
    Written by Advania Finland

    Advania has been helping Valmet to design and build a machine-learning chatbot. The chatbot is part of Valmet’s intranet, and it is designed to make it easier for staff to find and structure information. Neither Advania nor Valmet had any previous experience of the technology involved, and the project was a steep learning curve for both. Close cooperation was crucial for making the chatbot as efficient a tool as it is already proving itself to be.  

    Valmet’s staff rely heavily on the company’s extensive Microsoft-based intranet. However, with 14,000 employees around the world, ensuring that everyone has access to the information they need means that the intranet is overflowing with content. The need to do something about the content clutter arose when various kinds of challenges relating to the accessibility of information began to pop up in staff surveys. Valmet’s staff were, for the most part, very happy with the company’s intranet, but it had become laborious to search for and find information.   

    A leap into the unknown  

    It just so happened that, right around the time of the publication of the results of the latest staff survey, Valmet’s Head of Digital Communications Kimmo Miettinen was attending a seminar that focused on the use of chatbots in customer service. Miettinen was struck by the possibility of also using chatbot technology in an intranet environment to help in-house staff.  

    The idea began to properly take shape in 2019 when Valmet invited Advania to a tentative hackathon exploring chatbot technology. The goal was to test the concept and how it would work in practice.  

    ‘We had to choose between two options: one was to build on top of our existing SharePoint search engine, and the other was to design something completely revolutionary. We chose the latter option’, Miettinen explains.  

    Advania (formerly Blue Meteorite) has long been providing Valmet with, for example, intranet technology, and it felt natural to recruit a partner who already knows the environment to also build the chatbot. Advania’s solid expertise in Microsoft products also played a role in the choice of supplier.  

    ‘We named the tool “Charlie the Chatbot” and designed a sympathetic character for it in our intranet environment. Charlie is still young and learning all the time. We wanted to make the chatbot likeable as well to make the entire experience of using it to search for information an inviting one’, Miettinen says.  

    Machine learning helps to structure search hits  

    Valmet’s chatbot combines a SharePoint-based search function and a number of Microsoft’s AI functionalities. The chatbot runs on a machine-learning platform: it learns more every time a member of staff uses it and gradually starts to retrieve better and better hits. Miettinen’s team also includes a ‘chatbot whisperer’, whose job is to train the chatbot by feeding it pairs of questions and answers and monitoring staff’s satisfaction with the tool.  

    ‘Our staff have been instructed to rate their experience every time they use the chatbot, and the feedback will help Charlie to learn to retrieve more relevant hits’, Miettinen explains.  

    ‘Advania’s team were excited to get involved in the development of something as innovative as this. We all learned something new during the project. There were also some challenges, which we overcame by working together. Both parties now know more about artificial intelligence and machine learning’, Miettinen says.  

    Time savings and user-friendliness  

    At some point during the project it became clear that the chatbot was already finding better hits than the SharePoint-based search engine. That was the point when a decision was made to go live.  

    ‘Our staff have really embraced Charlie the Chatbot. We have already got a lot of feedback about the relevance of search results, but we still need more. The chatbot learns every time a user rates its performance, and we just have to be patient. I have no doubt that, in a year's time, Charlie will be awesome’, Miettinen says.  

    Further development in the pipeline  

    Valmet is already thinking about the future. Among the ideas floated so far is expanding Charlie’s territory to, for example, SharePoint workspaces, Teams and Valmet’s public websites. Another potential area of development is personalisation: the chatbot could be taught to identify individual users and retrieve information that is particularly relevant to them.  

    ‘Everything has gone extremely smoothly so far. Advania worked hard to make this a positive experience for us, and they certainly succeeded’, Miettinen concludes.  

     
    Share:

    Related articles