What challenges did Bevar Ukraine face before AI implementation?

Refugees needed conversational, multilingual, context-aware answers about housing, healthcare, and legal rights. Static FAQs and basic chatbots were insufficient. Data privacy and trust were critical requirements.

What was the impact of the AI solution?

Victor supports hundreds of users daily. In the first 150 days, the assistant saved 1,500 volunteer hours. The organization can support more people without proportionally increasing operational costs.

Source

ProductionEvidence: Medium50/100

Bevar Ukraine builds Victor generative AI virtual assistant with Amazon Bedrock for refugee support

Use case typeConversational supportUpdated Feb 5, 2026

Bevar Ukraine is an independent Danish non-profit humanitarian organization that supports displaced Ukrainians in Denmark with social and legal help. It built Victor, a generative AI virtual assistant to answer multilingual, context-aware questions about housing, healthcare, employment registration, and legal rights. The solution was developed with AWS using Amazon Bedrock, Amazon Titan Embeddings, Amazon EC2, and Amazon S3, with human-in-the-loop escalation and GDPR-focused security and privacy controls.

Organization: Bevar Ukraine
Industry: Other
Location: Denmark
Published: February 2026

Reported outcomes

1,500 hours

volunteer hours savedTime & speed

100 users/dayusers supported daily

Strategic outcomes

Cost efficiencyReduced routine volunteer workloadCustomer experience & trust24/7 refugee supportRisk & complianceBuilt privacy and trust into the architecture

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Volunteer hours saved: 1,500 hours decrease

AWS Public Sector BlogFeb 5, 2026Blog postExplicit claimMedium evidence strength

the assistant has saved 1,500 volunteer hours

Normalized claim

Users supported daily: 100 users/day increase

AWS Public Sector BlogFeb 5, 2026Blog postInferred claimMedium evidence strength

Victor now supports hundreds of users daily

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Bevar Ukraine
Provider: AWS
Maturity: Production
Linked source: AWS Public Sector Blog

The organization can support more people without proportionally increasing operational costs

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Conversational support
2Knowledge management
3Multilingual communication

Refugees needed conversational, multilingual, context-aware answers about housing, healthcare, and legal rights.
Static FAQs and basic chatbots were insufficient.
Data privacy and trust were critical requirements.

Developed Victor as a generative AI virtual assistant with Amazon Bedrock.
Used Amazon Titan Embeddings for semantic search over curated public-authority and social-service sources.
Designed a human-in-the-loop workflow so volunteers can review responses and refine the knowledge base.
Built GDPR compliance and data protection into the architecture from the beginning.

Technologies

Amazon Bedrock Amazon Titan embeddings Amazon EC2 Amazon S3 Claude 4.5 Sonnet

Victor supports hundreds of users daily.
In the first 150 days, the assistant saved 1,500 volunteer hours.
The organization can support more people without proportionally increasing operational costs.

Architecture

Victor is built on Amazon Bedrock with Claude 3.5 Sonnet, Amazon Titan Embeddings for semantic retrieval, and Amazon EC2 and Amazon S3 for application and storage. The architecture includes curated knowledge sources, human-in-the-loop review by volunteers, and security/privacy controls designed for GDPR compliance.

Sources & evidence1

Evidence: Medium50/100Evidence strength