We automate solutions to problems.
Located across three cities in Norway and Romania, we are a versatile team that focuses on improving business processes through automation. We build services based on machine learning algorithms and deploy them both internally and for external customers. Examples of our services include chatbots, auto-dispatching of tickets, and auto-recommendations of FAQ articles.
Our passion is artificial intelligence and we are constantly working to develop new services and improve existing ones, in order to add value to business. Our services reduce the repetitive, non-value adding tasks, saving time for our customers and allowing them to focus on more complex activities.
How we work
- The Method: Determines how we collect requirements, as well as design, build, test and release a service
- The Tools: Designates a set of approved digital tools that we use to build machine learning models, map out business process, and complete other important tasks
- The Environments: Establishes the environments (e.g., cloud) that customers can choose from when deploying the service
Our customers operate hundreds of applications, services and products on a daily basis. Customer satisfaction is at the core of what we deliver; any disruptions or incidents that affect the quality of our delivery must be dealt with quickly in order to maintain the trust and productivity of our customers. However, such incidents are impossible to eliminate completely. We therefore need to have strict procedures and rules in place as well as come up with smarter ways to prevent incidents from occurring in the first place.
To support operational teams in reducing incidents, the process automation team has produced a ”self-healing” service, which we call the Automation Engine. The engine provides automated incident handling—churning through large amounts of data to predict and prevent incidents. The service has been found to predict incidents with 87.47% accuracy and prevent them from happening 5 hours in advance. The result has been fewer incidents and more satisfied customers. In the future, the service will be implemented across the Visma Group and we see great promise in its use outside the company as well.
Service desks perform a vital function of resolving customer issues and requests. When these requests build up over time, they become backlogs, which reduce quality of service and lead to frustration for customers. They also affect the morale of those who work in the service desk itself: they work, work, and work, but it seems to never end. One of the main reasons for high backlogs is too many time-consuming manual tasks, such as dispatching tickets and searching for information to send to customers, like FAQ articles.
To address these backlogs, the process automation team developed a new service based on machine learning algorithms. Called Ticket Classification as a Service, or TCaaS for short, the service makes sure that incoming requests are automatically categorised and routed to the right person. The project’s success encouraged us to go a step further and improve the service with a new feature called DeeScovery. Through this feature, the robot recommends articles from the FAQ database, with an accuracy rate of 90%. It also analyses incoming requests, and recommends new articles that should be written to improve the database. Thanks to TCaaS and DeeScovery, service desk consultants can now offer better answers and have more time for other tasks.
In the process of debt collection, there are hundreds of activities Visma undertakes, many of which are done manually. Getting an overview of so many activities, undertaken by different teams at different times, is a challenging task when attempted by humans alone. But how can processes be improved when a full overview does not exist?
To address this need, the process automation team employed process mining—the set of techniques, methods and algorithms that perform analysis on log data. The objective was to visualise each step of the debt collection process and provide metrics—such as duration and distribution across the team—to help users better understand processes based on factual data.
Since the service was implemented, users have now attained a better understanding of the true volume of activities, which are most used, and how they are connected. The service not only helps them improve their processes, but also makes recommendations on how to improve the process, for example by suggesting the best next step.
Some customers use what is known as the Service Intensive Rate, or SIR, to gauge their performance in reducing customer requests. The SIR is the ratio of requests to licenses, and a good SIR is about 0.25. In other words, the goal is for fewer than 1 in 4 customers to submit a request in a given year.
One way to reduce SIR is through chatbots. We developed a chatbot that answers FAQs, redirects customers to relevant information, and even predicts their needs and comes up with answers before questions are raised. The chatbot was implemented in 2017 and after just one year handled 51% of all chats.
We conduct a quarterly market analysis of the service across key categories, including natural language processing, context awareness, learning from limited examples, system integrations, Nordic language support, and more. Our analyses show that this service is the most feature-rich, user-friendly, and cost-effective solution for our customers.
Want to join the team?
We operate in small teams with varied expertise in python programming, RPA development and machine learning. After implementation, we have skilled people managing the bots’ daily operations and maintenance. We are looking for people with the right attitude, meaning they are always willing to learn, and who love data, machine learning and all the great stuff AI can do. If you want to work in a cool, open-minded, highly professional team, please send an email to:
Oslo, Norway / Timisoara and Sibiu, Romania