Retrieval-augmented generation looks deceptively easy in a demo: upload some PDFs, ask a question, get an answer. The hard part starts with your real documents, your permissions and your accuracy bar. If you are shortlisting a RAG development company, ask these nine questions.
The nine questions
How will you measure accuracy on my questions?
The right answer involves building an evaluation set from questions your team actually asks, with agreed pass thresholds. "It works great in our demos" is the wrong answer.
How do you handle my document types?
Scanned PDFs, tables, slides, long contracts and wikis each need different parsing and chunking. Ask to see how they handle your worst documents.
How is retrieval tuned?
Look for hybrid search (keyword plus semantic), re-ranking where it helps, and metadata filters, not just a default vector search.
Does every answer cite its source?
Citations let your team verify answers and are essential for legal, compliance and customer-facing use.
What happens when the answer is not in the documents?
A good system says it does not know and escalates, rather than inventing something plausible.
How are permissions enforced?
If different people may see different documents, retrieval must respect that. Ask how access controls mirror your existing ones.
Where does my data go?
Ask which providers see your content, retention terms, and whether the system can be deployed in your own cloud account.
What will it cost in total?
Build cost, plus running cost (model usage, hosting), plus maintenance as documents change. A fixed build price removes one unknown. See how much RAG costs and when fine-tuning is cheaper.
Who owns it afterwards?
You should receive code, prompts, the evaluation suite and a runbook, not a black box tied to the vendor.
Red flags
- No evaluation plan, only a demo.
- One generic pipeline for every client.
- Vague answers on permissions and data residency.
- Hourly estimates with no definition of done.
New to the topic? Start with What is RAG and why your business needs it. For how we scope and build, see our RAG consulting and development service.
Frequently asked questions
How do I choose a RAG development company?
Ask how they will measure accuracy on your own questions, how they handle your document types and permissions, whether answers cite sources, where your data goes, and what you own at the end.
How much does it cost to hire a RAG development company?
It depends on the number and quality of content sources, permissions, accuracy bar and integrations. A reputable vendor will quote a fixed price after scoping and explain the running costs separately.
How long does a RAG project take?
A single-source RAG system can reach production in about 2–4 weeks. Multi-source or regulated-data systems take longer and the timeline should be agreed in writing.
Related: RAG development services · What is RAG?
Want a straight answer for your project?
Book a 30-minute call. I will tell you what I would do, and a fixed price if it makes sense.