Using AI to help 10,000+ refugees navigate Danish life with Bevar Ukraine

In 2022, after Russia’s full-scale invasion of Ukraine began, an influx of Ukrainian refugees arrived in Denmark, and Bevar Ukraine’s center for social and legal assistance was overwhelmed. “Refugees arrive in a new country, and they need to know how things function,” says Taras Tsarenko, a volunteer at Bevar Ukraine. “They need social and legal aid on a variety of issues.” At the time, Bevar Ukraine had only one employee who provided legal consultations, and he was often booked weeks in advance.

Many of the newly arrived refugees had the same questions about life in Denmark—how to find housing, enroll their kids in school, apply for a job, file taxes, book a doctor’s appointment, or find Danish language courses, for example. Bevar Ukraine decided to create an AI chatbot that could answer such questions in Ukrainian. The team collaborated with AWS to build the chatbot, and the effort benefited from AWS Promotional Credit through the AWS Nonprofit Credit Program, which helps offset costs for nonprofits associated with implementing cloud-based solutions. “The AWS team members were very helpful and saw the value of the project,” says Tsarenko, who headed the development. A handful of Bevar Ukraine volunteers—Tsarenko, Vitalii Bozadzhy, Vladyslav Horbatenko, Katya Tkachuk, Ruslana Onyshchuk, and Martin Vyuga—worked for 3 months to make the first release of the chatbot a reality.

Bevar Ukraine’s AI-powered virtual assistant, Viktor, went live in 2025. It uses a knowledge base of over 400 official Danish resources to provide responses to questions about Danish life, legal systems, and social services. “People were living their lives, they leave Ukraine, and they arrive here totally confused,” says Tsarenko. “There’s a lot of complexity and ambiguity, everything is new, and when you have someone to chat to, you can feel supported as you face this tough reality.”

One Ukrainian used Viktor to find out what to do after a minor car accident. In another case, the 60-year-old mother of one of the developers expressed her appreciation at being able to communicate with the chatbot in her own language. “Understanding life in Denmark can be tough, but when Viktor started to answer her simple questions in Ukrainian, she was very happy,” says Tsarenko. “It’s not just about solving a problem but about having someone guide you through your story.”

The chatbot evolved from using a single large language model (LLM) and retrieval augmented generation architecture to become a complex agentic AI system. Bevar Ukraine used Amazon Bedrock—a solution for building generative AI applications and agents at production scale—to develop around 150 specialized AI agents. Those include translation agents, agents for legal and healthcare expertise, and orchestrator agents, which break down questions to produce more precise answers. This process of breaking down the question is especially helpful for users who may struggle even to express the problem or question that they have.

As it developed Viktor, Bevar Ukraine followed the AWS Well-Architected Framework, which helps an organization understand the pros and cons of decisions it makes while building systems on AWS. The organization’s lawyer also worked with the IT team to verify that the solution complied with European regulations and to vet the resources in Viktor’s knowledge base. The team tested the chatbot for several months before releasing it more widely. The chatbot, which originally spoke Ukrainian, Danish, and English, is now accessible in over 100 languages.

Over 10,000 people have used Viktor since its launch. This includes 4,000–5,000 monthly active users, with an average of seven requests made per person.

Around 100 people use Viktor 10–15 times per day, likely social workers or municipal employees. “We sent information about Viktor to all the municipalities in Denmark, and many of them use the service to advise local Ukrainian refugees,” says Kuzmyn. “It has not only freed up our own resources but also municipality resources.”

The chatbot has answered over 100,000 questions in the first year, far more than a single employee could have managed. The integration advisor at Bevar Ukraine can now dedicate his time to more complex cases. “Previously, people sometimes needed to wait weeks for very simple questions to be answered,” says Tsarenko. “Now, if you have a simple question, you can get an answer right away, 24/7. If you still need to talk to someone in person, you can at least know you are asking the right questions because Viktor helps you understand the situation.”

Bevar Ukraine hopes to expand the chatbot to help other refugees in Denmark, as well as Ukrainian refugees in other countries. “Because we have a multi-service architecture on AWS, we can scale the system and replicate the technology behind the chatbot very simply,” says Tsarenko. “We’re now looking for collaborators to help us adapt Viktor to other countries.”