Victor AI Multilingual help for refugees in Denmark.
A help service that gives refugees practical guidance on public services, healthcare, housing, work and education, in their own language, powered by more than 100 specialized AI agents.
Developed in collaboration with AWS
Capabilities
Two things make the architecture work.
The case
Refugees arriving in Denmark have to find their way through public services, healthcare, housing, employment, education and legal procedures. Information is scattered across authorities and websites, usually in Danish or English. Support organisations cannot give personal guidance to everyone.
Victor AI, developed in collaboration with AWS, is a multilingual agentic platform. People ask in their own language, from over 100, and Victor routes each question to one of more than 100 specialised agents, each focused on one information need. Questions that need more attention go to a person.
Guidance is accessible whatever the language, and it scales across domains without scaling the support team. Routine questions no longer fill staff time, so people can concentrate on the cases that need them. The architecture can take on new services as they are needed.
How it works
From a question in any language to the right agent.
The same path serves every topic. What changes is the agent that answers.
- 01
Ask
A person writes in their own language, in plain words.
- 02
Language layer
The platform detects the language and translates in both directions, so nobody has to switch to Danish or English.
- 03
Route
A router identifies what the question is about and selects the right specialised agent.
- 04
Specialised agents
Each agent covers one information need, such as healthcare, housing, employment, education or public services. New agents extend the platform without redesigning it.
- 05
Answer
The reply comes back in the user’s own language as practical guidance.
- 06
A person steps in
When a question needs more attention, it is passed to the support team.
Value
Business and social value.
- Accessible guidance, regardless of language.
- Scalable assistance across many information domains.
- Less repetitive work for support organisations.
- More efficient use of human support resources.
- An extensible architecture for future services.
We do not publish adoption figures or results for this case.
Concept
A grant manager for NGOs.
The same pattern — specialised knowledge, clear sources, a person at every decision — applies to another bottleneck in non-profits: finding, writing and reporting on grants.
- 01
Calls and profile
Funder calls are collected and compared with the organisation’s mission, history and capacity.
- 02
Match and eligibility
Opportunities are ranked, and each eligibility judgement cites the rule it relies on.
- 03
Draft
Sections are drafted from past applications and project data. Figures are sourced, never generated.
- 04
Grant manager decides
A person edits and approves every section. Nothing is submitted automatically.
- 05
Report and remind
Reports are assembled from project records, and a tracker keeps deadlines and obligations visible.
Concept status
The grant manager is a concept CoreAI would explore with an NGO. It is not a delivered system, and no results are claimed.
Contact
Ready to make AI operational?
Whether you need an AI roadmap, a reference architecture or a working prototype, let’s discuss where to start.
- contact@core.ai
- linkedin.com/company/coreai-dk
- Location
- Copenhagen, Denmark