Skip to main content

Using artificial intelligence to solve problems for our customers

Artificial intelligence and machine learning represent new opportunities and a fundamental change in the way we work. Find out how we use these technologies to create value for our customers.

Visma helps Nordic governmental giant take the next steps within AI

Artificial intelligence (AI) is sometimes described as one of the biggest opportunities to solve the problems of the world and support the growth of the wellbeing of human beings. At the same time, some are talking about AI as a huge threat.

One part of artificial intelligence is machine learning (ML). Visma is utilising machine learning in products and more than 200.000 customers of Visma have access to functionality where machine learning is adding value.

Visma believes that AI/ML can increase value for Visma’s customers through automation of processes, improved usability, increased insights or knowledge and increased quality.

Opportunities

One of the easiest ways of utilising machine learning is the classification of data. Some examples of classification are recognising information in a picture or assigning an incoming task to the right person or group based on information in the task. This methodology is used to propose booking of supplier invoices, propose who to approve a supplier invoice, assign support ticket to the right group or to propose knowledge articles to a customer in touch with support. Over time, high-quality classification of data can transform processes.

A side-effect of automation is reduced user errors while typing. Machine learning and statistical methods can also contribute to improvement in data by identifying deviations and either automatically correct them or ask a human being. Such methods can also be used to prioritise sales prospects and identify potential fraud.

Currently, most software is equal to all users unless customisations have been applied. Machine learning can be used to analyse patterns of usage and predict the next step for a user or even to guide the user when she/he is in doubt. Speech-based interaction with software simplifies the usage in some situations. Chat-based interaction with a system is in some situations easier than using menus, links and buttons. Speech-recognition and chat-bots to understand language and intent are heavily using machine learning.

Predicting the future based on large amounts of data is another potential area of usage of machine learning. Small businesses are often not making detailed budgets and predictions of the future based on history combined with easy to understand analysis might be a better tool for strategic decisions.

Value for Visma customers

The 200.000 customers of Visma using a product including machine learning are getting value through proposals of nominal codes, matching of items in bank reconciliation and proposals for how to book a supplier invoice. About 2 million proposals are given each month to the users of our systems. In addition, Visma is using machine learning to interpret pictures of receipts and invoices which means less manual typing for our clients. Some customers have access to chatbots to get help about products around the clock.

Currently, Visma is testing machine learning and statistical methods in several projects to increase value for customers. Machine learning requires a lot of data to be feasible and we have started with the processes where we or others are having enough data. Based on technology from others Visma is testing both chat-based interaction with our ERP-systems to enable users to write questions like “How much was last months sales to customer <NN>” to get information from the system or to type in commands like “create invoice for customer <NN>”. Another way we are testing out new interaction methods is to add speech-recognition on top of our public APIs to be able to talk natural language to the system.

Several product-teams are using analysis of usage to improve the usability of the service and are playing with models for adapting the UI to each user. Some teams are working on using ML to classify data to reduce the need for input from the user.  Another project is testing out how we can generate a short article/text describing the same information as in a profit & loss -report including describing the most important trends and deviations. One team is working on predicting the revenue and profit in the next months of a small business. Our security- and platform team is working models for intrusion detection and other security-related issues.

All these projects require changes to the current software and Visma is updating the software regularly. At the same time, some customers are trying out standardised software robots to operate some of the installed software of Visma to automate and simplify.

What all these projects have in common is to solve actual problems for our customers. The only thing that is sure is that we will have even more ideas in the future. Some of the things we are working will not work due to the too low accuracy or too little data available while others will be available for customers.