Skip to content

FAQ

No silly questions. Just clear answers.

About Vitinn

  • Is it safe to use sensitive company data with Claude or ChatGPT?Security & privacy

    It depends on where and how the models run. With public subscriptions, the data goes to a service you don’t control, and the terms decide what happens to it.

    In Vitinn, the models run in your own cloud environment in Europe. The data is encrypted, you hold the encryption keys, and what people put in is not used to train models.

    More on security

    Was this answer helpful?
  • What is APRÓ Vitinn?AI at work

    A secure AI workspace that runs in your own cloud (AWS or Azure) in Europe and is built on your own data and ways of working, not just general information.

    Your people get a chat interface connected to your internal systems. You choose the models and own the data and the solution, and we help you roll it out.

    More about Vitinn

    Was this answer helpful?
  • How is APRÓ Vitinn different from ChatGPT?AI at work

    The main difference is security. Data that goes into Vitinn, or is created there, is accessible only to you, with your own encryption keys, and you can be certain it’s stored in the EU.

    On top of that, Vitinn runs in your environment, uses your sign-on and connects to your data. The price is fixed per month, not per user.

    Was this answer helpful?
  • How is Vitinn different from Microsoft Copilot?AI at work

    They don’t rule each other out. When Vitinn runs in Azure, it is built on Microsoft Foundry and runs in the same Microsoft environment, so many companies use both.

    Copilot is built into Microsoft 365 and is at its best when the work happens in Word, Outlook, Teams and Excel. It is paid for as a licence per user.

    Vitinn runs in your own environment. It suits you when:

    • you want to choose from more models, Gemini or others for example, and use whichever suits each task best
    • many people need access, contractors too - there is no licence fee per user, you pay for usage
    • the data sits in systems outside Microsoft 365
    • personal data should be filtered before it reaches the model
    • you want to see what each project costs and what it delivers

    Copilot for the work inside Office. Vitinn for everything else.

    Was this answer helpful?
  • How does Microsoft Foundry compare with Vitinn?AI at work

    Copilot is ready-made in Office. Foundry is a workshop. Vitinn in Azure is a foundation built in Foundry, ready to use.

    Copilot is a no-code solution. It works in Word, Outlook, Teams and Excel, and in Copilot Studio you can put together simple agents without code. Microsoft Foundry is the environment where developers build AI solutions in code - models, integrations, access control and monitoring, all built and maintained by a development team.

    Vitinn brings the two together. In Azure it is built in Foundry, but the foundation is ready: an interface for staff, access control, filtering of personal data and an overview of cost. You start with the projects, not the foundation - and when something custom is needed on top, it is built in Foundry like any other Foundry project.

    Was this answer helpful?
  • What does Vitinn cost?AI at work

    A fixed monthly price based on scope, with no per-user subscription.

    The basic setup is included; model usage and custom builds are paid separately. Model usage is paid directly to AWS or Azure on your own account. You see it and control it in Vitinn.

    Get a quote

    Was this answer helpful?
  • Who pays for using the models?AI at work

    Model usage is paid directly to AWS or Azure on your own account. You see it and control it in Vitinn.

    Was this answer helpful?
  • How long does it take to set up Vitinn?AI at work

    It depends on the scope and the integrations. We go through it in the first conversation.

    Book a Consultation

    Was this answer helpful?
  • What happens if we stop?AI at work

    No vendor lock-in: the environment and the data are yours and stay in your cloud environment.

    Was this answer helpful?
  • Which AI models can we use?Tools & APIs

    At setup you choose which models are available, whether Claude, ChatGPT, Gemini or others. You can add or switch models as the technology moves on, without starting over.

    Was this answer helpful?
  • Can we control usage and cost?Security & privacy

    Yes. Usage, cost and access to models are managed centrally, for example with a budget and rules you set.

    Was this answer helpful?
  • Who has access to the system, and can that be changed?Security & privacy

    We use your access control. Most often that’s Microsoft Entra single sign-on (SSO), but Google SSO works too.

    Behind the scenes, only authorised users can reach the data and systems they’re allowed to, and flexible permission management allows fine-grained settings. You can change access at any time.

    Integrations with internal systems

    Was this answer helpful?
  • How is my data protected? Will it be used to train other models?Security & privacy

    All data is encrypted, in transit and at rest, and you have full control of the encryption keys. Data is never shared and never used to train the underlying models.

    Security and data

    Was this answer helpful?
  • Does the service follow common security standards?Security & privacy

    Yes. The setup is aligned with international security standards such as ISO, SOC and CSA, and meets the requirements of HIPAA and European data protection law (GDPR).

    APRÓ, together with AWS and Azure, provides detailed support to help you meet compliance requirements for your own operating environment.

    Was this answer helpful?
  • We have little experience with AI. Where do we start?AI at work

    We offer both advice and support for rolling out Vitinn. You can also get a tailored interface that suits your workplace, with the relevant data made available through it.

    See use cases

    Was this answer helpful?

About AI

  • What is an agent?AI at work

    An agent is AI that does a task rather than just answering questions: it fetches data from systems, follows steps and delivers a result, within the permissions it has been given.

    Was this answer helpful?
  • Is using AI at work cutting corners?AI at work

    At APRÓ we see it the other way round. AI is one of the most remarkable tools of recent decades, making both simple and complex work faster and easier. But it’s first and foremost a tool that strengthens the skills and judgement your people already have.

    That’s why a clear policy and guidelines matter. Secure access to AI prevents shadow AI, data leaks and other operational security risks.

    Was this answer helpful?
  • Can AI create images and audio?Image & audio

    Yes, but usually with different kinds of models from the ones that write text. Image models create pictures from a description, and speech models turn speech into text (speech recognition) or text into speech (speech synthesis).

    Language models such as Claude work mainly with text. They can read and describe images, charts and documents, but don’t create images themselves.

    Was this answer helpful?
  • Does AI understand spoken Icelandic?Image & audio

    Speech recognition and synthesis for Icelandic have improved a lot in recent years, helped by the government’s language technology programme. Quality depends on the model and the audio: clear speech in a quiet room works far better than a recording of a noisy meeting. It pays to test on your own material before settling on a tool.

    Was this answer helpful?
  • What is an API, and why does it matter for AI?Tools & APIs

    An API is a way for systems to talk to each other. AI that can reach your systems through an API can fetch information and do work, such as looking something up in the books or creating a ticket, instead of answering from general knowledge alone.

    Was this answer helpful?
  • What is MCP?Tools & APIs

    MCP (Model Context Protocol) is an open standard for connecting AI to systems and data. Each connection describes the tools it offers, and the AI picks and uses the right ones based on what you ask. Because the standard is open, the connections come with you if you later switch model or interface.

    Vitinn’s integrations

    Was this answer helpful?
  • Can AI help with programming?Tools & APIs

    Yes. Language models are good at explaining code and error messages, drafting functions and tests, and documenting what exists. The developer stays responsible for the code: they review it, test it and decide what ships.

    Was this answer helpful?

Got a question about AI that isn’t here?

We’ll answer you by email, and add the answer here if it’s useful to others.

Questions we’re working on

  • How do I run a local model without sending data to the cloud?
  • Which tool works best for Icelandic speech recognition today?