LLM questions, answered
Short answers to the questions business and IT leaders ask us most often about large language models.
Is it safe to use ChatGPT-style tools with company data?
It depends on the service and contract. Consumer tools may keep or use data in ways you have not agreed to. Enterprise offerings and private deployments give you control over retention and location. Give staff an approved option, or they will use unapproved ones.
Will an LLM make things up?
Yes, sometimes. Grounding answers in your documents with RAG, asking for sources, and allowing "I don't know" reduce it a lot. For important decisions, a person should check the output.
Do open models work in Hindi and other Indian languages?
Some do well, some do not, and quality varies by language and task. Test with real examples from your users before choosing.
Hosted API or our own GPUs?
Hosted APIs are usually cheaper at low and medium volume and give the strongest models. Self-hosting makes sense for strict data rules or steady high volume. Use the cost calculator with your numbers.
Do we need to fine-tune a model on our data?
Usually not at first. To answer from your documents, use RAG. Fine-tune later for format, style or narrow repeated tasks. See fine-tuning.
How long does a first production use case take?
Typically eight to twelve weeks: test set and model comparison, a guarded build, and a pilot with real users. The test set is the step most often skipped, and the one that matters most.
More LLM guides: Choosing a model · Private hosting · Fine-tuning · Governance and LLMOps · LLM FAQ · Use case: claims summaries
Starting with LLMs, or stuck after a pilot?
Tell us the task you want AI to help with and any rules about where your data can go. We will come back with a plain recommendation: which kind of model, where to run it, roughly what it costs, and how to know if it is working.