Frequently Asked Questions
Answers to the most important questions about AI projects for mid-sized companies.
AI Projects
What industries does bitvaria work in?
We're industry-agnostic. Our expertise lies in the technical implementation of AI applications, not in a specific industry. Our three focus areas (unlocking knowledge, automating processes, building custom applications) are relevant across industries.
What AI applications does bitvaria develop?
We focus on three areas: Unlocking knowledge (making company knowledge searchable), automating processes (shortening manual workflows with AI), and building custom applications (AI software that fits your processes). These focus areas cover most practical use cases in mid-sized companies.
How long does it take to develop an AI application?
Depending on complexity, anywhere from a few weeks to several months. A simple knowledge assistant can be production-ready in 4-8 weeks. More complex applications with many integrations take longer. We'll give you a realistic estimate after the first conversation.
What does an AI project cost?
The entry point is the platform pilot: 8,900 EUR net, three months, one department, running in your own cloud account. It is credited in full against the full rollout. A custom AI application as a project starts in the mid five-figure range, depending on scope and complexity. An initial conversation is free. We work with transparent budgets and regular sync meetings. No nasty surprises at the end.
Does bitvaria use existing AI models or train custom ones?
We use existing models. Training your own from scratch is neither necessary nor economically sensible for the vast majority of applications; quality is decided in the application around the model, not in the model itself. Which model does the work depends on the task: everyday work like summarizing, classifying or drafting runs well on open models in European data centers, harder cases on the large commercial ones. Which of them leads changes every few months, so we develop in a way that keeps a model swap a matter of configuration rather than new development. What belongs to you is unaffected by any of it: your data, your integrations, the rules governing which task may go where, and the tests that let you evidence quality.
Collaboration
What does a typical project look like?
We start with a conversation to identify your potential. Then we jointly design the solution: what it should do, what it looks like, what it needs to connect to. During development, you see regular progress updates. After launch, we stay on it and continue developing what proves valuable in daily use.
Does bitvaria offer maintenance and support?
We do not offer classic maintenance or support contracts. We build your application to run independently and hand it over, source code and configuration included, to your team or your IT provider. One exception: for the AI platform we keep assessing the model market and re-measure quality, because something changes there every few months. That is recalibration at a fixed price, not a standby service.
What does collaboration look like in practice?
You keep the same point of contact throughout, someone who knows your project from start to finish. Regular sync meetings (usually weekly) ensure you always know where we stand. We use modern tools for project management and communication. You decide what works for you.
Technical
What systems can be connected?
Basically any system with an API: ERP systems (SAP, Microsoft Dynamics, etc.), CRM systems, databases, cloud services. For older systems without modern interfaces, we usually find ways too, e.g., via file exports or middleware.
Where are the applications hosted?
In your own cloud account, not in ours. Technically that usually means AWS, the platform we have worked with longest. The difference is not the technology but who owns the accounts: you keep access, billing and control, even if you terminate the engagement. Hardware in your own building is possible, and it is the right call when data must not leave the premises for regulatory reasons. As a cost-saving measure it rarely pays off.
How secure is the data?
Data security is not an afterthought, but part of the architecture. We work with encryption, access controls, and audit logs. For sensitive data, we can also implement solutions without data transfer to external AI providers. We'll tailor the specific security architecture to your requirements.
Question not listed?
Write to us. We typically respond within one business day.