Today we will dive into a customer case study. Thanks to the Visma team on their amazing partnership!

Visma Spcs, is a leading software company providing modern services to simplify accounting, HR, payroll, and more, and they’ve successfully integrated AI and Semantic Kernel to revolutionize their customer experience. With hundreds of thousands of customers across the Nordic region, Visma Spcs recognized the need to empower entrepreneurs and support accountants in achieving their goals.

“Improving customer experience is key,” affirms Roger Andersson, Head of Technology. “AI, powered by Semantic Kernel, helps us address these challenges.”

Understanding Customer Needs

“We at Visma Spcs empower entrepreneurs to pursue their dreams and enable accountants to support their clients in their pursuit of achieving their goals. For us, it boils down to building the right things in the right way, and thereby creating value for our users. This is where AI enters the equation”, says Product Experience Manager Johan Björnegård.

After conducting customer research, we have identified the following common needs among our customers:

• Finding the right information quickly and easily
• Asking questions and getting accurate answers
• Locating references to the correct documents and information

Addressing Customer Needs with AI

We at Visma Spcs wanted to improve the customer experience on the above points, says developer Jimmy Stridh. What we have done so far is to integrate AI and Semantic Kernel in using Retrieval Augmented Generation (RAG) to provide insightful answers over our existing product documentation. For this we’ve also utilized Azure OpenAI with GPT-4, and Azure AI Search. This gives us powerful tools to build on the existing RAG solution, as well as a clear path forward to where we’d like to take the user experience.

On Semantic Kernel

We chose Semantic Kernel as a solution as we are an organization that primarily uses .NET in our tech stack, and this provides smooth adoption throughout the company. We especially appreciated the following features:

• The extensive support and development around orchestration, agents and automation. We are deeply integrating our generative AI solutions in our products which means this is a vital area where we also see Microsoft and Semantic Kernel leading the way for developers.
• Abstraction and flexibility over AI model. As the landscape develops and specialized models appear along with new better foundation models, we have the option to select the correct model for every use-case.
• The proven track-record of Microsoft in developing copilots show their dependability as a partner in the ongoing evolution of this field.

Additionally, we appreciate the efforts being made on the framework, and the development of the agents moving forward – something that has become increasingly important as we have started new projects in the area.

Delivering in Real-Time Customer Interactions

To date, the AI-improved solution, called the AI Assistant, has been rolled out to hundreds of thousands of customers and the feedback and metrics has so far been very positive.

We’ve measured the following metrics to assess the performance of the AI solution:

• Usage rate: Several percent of daily active users interact with the chat daily, and we see a lot of recurring users.
• Correctness: Close to 90% of chat requests are resulting in a good response for the end-user, and trending upwards. Like many, we have seen the importance of having well-documented product information as a prerequisite for good performance.
• Latency to answer: Steady at a couple of seconds to first response. We observe that the ebb and flow of latency throughout the day, heavily depends on worldwide API usage more than anything.

The above metrics made us confident that we were on the right track and that the solution was working as expected.

What does the Customer Say

The feedback from customers has been overwhelmingly positive. They appreciate the convenience of receiving support outside of regular office hours, with around 40% of messages coming in when human support is unavailable. This 24/7 availability has been a significant benefit, allowing customers to get assistance whenever they need it most.

Furthermore, the AI Assistant has proven to be a valuable resource for the customer success team, particularly for new team members. When faced with customer inquiries in areas they are still learning about, the AI Assistant provides guidance and support, helping them navigate these challenges more effectively.

Semantic Kernel Team and their thoughts on the solution from Visma Spcs

“I was very impressed with the solution that Visma has implemented. It’s great to see how they have been able to use Semantic Kernel to improve the customer experience. It’s positive to see they’ve leveraged Semantic Kernel and the power of AI in their AI Assistant with hundreds of thousands of customers with 90% accuracy” says Matthew Bolanos.

What’s Next

After the initial success and rollout, we’re looking to improve the solution even further. We’re looking to do the following in the next phase:

• Create tight integrations into our applications, utilizing plugins and agents to enable action-taking from the AI Assistant.
• Employ the planner and agents concepts of Semantic Kernel to give a new dimension of the AI Assistant giving our services a new interface for the customer to interact with.
• Rollout to our full product suite containing dozens of products.

Summary

Choosing Semantic Kernel to enhance our customer experience has proven to be a great decision, providing significant benefits such as abstraction over large language models (LLMs), which allows us to select the most appropriate model for the task at hand. Additionally, its powerful orchestration capabilities have just begun to unveil their potential, suggesting that their importance will only grow as we continue to explore and utilize these features more extensively.

We’ve also appreciated being able to share our solution and experiences with PM Matthew Bolanos and the Semantic Kernel team as the rapid progress in the development of Semantic Kernel continues.

References

• Learn more about Visma Spcs and their offers here: Visma Spcs 
• Learn more about Semantic Kernel: GitHub Microsoft Semantic Kernel, Semantic Kernel documentation | Microsoft Learn

Please reach out if you have any questions or feedback through our Semantic Kernel GitHub Discussion Channel. We look forward to hearing from you! We would also love your support, if you’ve enjoyed using Semantic Kernel, give us a star on GitHub.