Question 1

Isn’t ChatGPT enough?

Accepted Answer

For generic questions, yes — but for grounded answers on your data without hallucination, you need RAG and custom business rules.

Question 2

Which models do you use?

Accepted Answer

OpenAI (GPT), Anthropic (Claude), Azure OpenAI, open-source (Llama). We pick or combine based on the need.

Question 3

Where is the data stored?

Accepted Answer

Turkey / EU servers, OpenAI no-train option, or your own infra (on-prem). KVKK/GDPR ready as standard.

Question 4

What about cost?

Accepted Answer

Per-conversation cost is modelled; we use a fast/cheap model as a first filter and a larger one as a fallback to optimise the budget.

Question 5

Will we replace staff?

Accepted Answer

No — the assistant covers repetitive and after-hours work. Your team focuses on complex, high-value cases.