Frequently Asked Questions
Answers to the most common questions about AI consulting, data security, model selection, custom AI solutions and training. If you cannot find what you are looking for, write to us.
General
Who is Huaris AI?
Huaris AI is the AI consulting brand of Huaris Technology. We are a vendor-independent consultancy offering AI strategy, model selection, custom AI solutions, security and compliance, and team training to organisations.
Which AI providers does Huaris AI work with?
We are not tied to a single provider. Depending on the need we recommend cloud models such as those from OpenAI, Google (Gemini) and Anthropic (Claude), or open-weight models such as Meta Llama and Mistral. We do not claim to be an official partner of these companies.
Can I buy the services separately?
Yes. The five services can be taken on their own or combined; for example a strategy engagement followed by a custom AI solution and team training. We define the scope in the discovery call.
How do I get started?
Write to [email protected] to request a free discovery call. In the call we listen to your processes and goals and evaluate together where AI can add value.
What is the relationship between Huaris AI and Huaris Technology?
Huaris AI is the AI consulting brand of the Huaris Technology group. You can learn about the parent company at huaristech.com.
Data security and KVKK
What happens if company data is sent to tools like ChatGPT?
It depends on the provider and the plan. For individual plans, whether content is used to improve the product depends on the settings and the provider's terms. For enterprise and API offerings, providers generally state that data is not used for model training by default.
We verify each provider's current terms with you at the contract stage.
What should be considered under KVKK when using large language models?
If personal data is processed, topics such as the duty to inform, legal basis, data minimisation, data processing agreements and cross-border transfer come into play. Where possible, architectures that never send personal data to the model are preferred.
This is a general overview and does not replace legal advice.
Does Huaris AI issue a legal compliance certificate?
No. We design the technical architecture and documentation with regulatory requirements in mind; legal assessment and any formal compliance decision belong to your legal advisers.
Will an on-premise model be as good as a cloud API?
Quality depends on the model and the task. Open-weight models can be sufficient for many enterprise tasks, but may differ from the most advanced cloud models. We decide based on an evaluation with your own data.
Related service: Product Development, Security and Compliance
Model selection and technology
Should we choose a cloud API or an on-premise model?
Data sensitivity, regulatory requirements, budget, expected quality and operating capacity decide. When data must not leave your organisation, an on-premise open-weight model is usually the choice; otherwise an Enterprise API is generally more practical.
Why vendor-independent advice?
We make recommendations independent of any vendor, based only on your needs. Models change quickly, so what fits best today may change tomorrow; that is why we recommend an architecture that can be swapped where needed.
Should we use one model or several?
For some tasks one model is enough. If tasks have different quality and cost requirements, using several models together can be more efficient. We decide based on evaluation results.
Related service: Model Selection and Vendor-Independent Advisory
Custom AI and RAG
What is RAG?
RAG (Retrieval-Augmented Generation) means the model first retrieves relevant passages from your own documents and grounds its answer in them. Answers can cite sources and the risk of hallucination is reduced, but not eliminated, which is why we run evaluation tests.
Setup time depends on data volume and integration scope; we clarify it in the discovery call.
Does RAG completely prevent hallucination?
No. Grounding answers in your sources reduces the risk of hallucination but does not eliminate it. That is why we set up source citation, behaviour where the assistant says when it does not know, and regular evaluation tests.
Does our data go to the model and outside?
It depends on the architecture. With an Enterprise API only the document passages relevant to a question may be sent to the model; if privacy requirements are high, the model can run inside your organisation (on-premise). In every case we document the data flow.
Related service: Custom AI Solutions
Strategy and discovery
When does a discovery engagement make sense?
It makes sense before starting an AI investment, or when earlier experiments did not deliver. The goal is to start from business problems rather than technology and find the few areas with the highest value. Duration and scope depend on the size of the organisation; we clarify them in the first call.
How is the ROI analysis calculated?
We calculate items such as working time saved, cost of errors and rework, and build and running costs using your own data and explicitly written assumptions. We also show how the result changes when the assumptions change.
Related service: AI Strategy and Discovery
Team training
Which tools does the training cover?
It is shaped around the tools you use or plan to use, for example Copilot, ChatGPT or custom assistants. We also teach tool-independent prompt writing principles.
What is a prompt library?
It is a collection of tried, reusable prompt templates for tasks a department performs often, so everyone starts from consistent starting points instead of from scratch.
Related service: Team Training and Prompt Engineering
Have a question?
If you did not find the answer you were looking for, write to us and we will get back to you.
Send an email