AI Support That Cites Its Sources
An accuracy-first AI support assistant for SaaS. Every answer cites its source and it says “I don’t know” instead of guessing — built for teams who can’t afford a confidently wrong bot.

Grounded — designed, built & operated by Lumina
Accuracy-First AI Support Assistant
SaaS teams who can’t afford a confidently wrong bot
Automate support answers without the bot ever guessing or inventing an answer
Most AI support bots hallucinate — answering confidently even when they’re wrong.
There’s no way to see where an answer came from, so no one trusts it.
Chasing a high “resolution rate” pushes bots to guess instead of admitting uncertainty.
We designed and built an accuracy-first AI support assistant that answers from a team’s own documentation, cites the exact source for every answer, and refuses to guess when the answer isn’t there. It’s a product we design, build, and operate end-to-end.
Three interlocking layers make it reliable: a retrieval layer that finds relevant passages, a generation layer that stays inside those passages, and a refusal layer that fires when there’s nothing good to say.
Docs are chunked and embedded into pgvector. Hybrid semantic + keyword search finds the right passages for each query.
Answers are written only from retrieved passages, each tied back to the exact source section it came from.
When retrieval comes up short, the assistant says “I don’t know” instead of guessing — by design, not by accident.
A working support assistant that answers from a team’s own docs and links every answer to the exact source
Honest behaviour by design: it says “I don’t know” and refuses to guess when the answer isn’t in the docs

You’d work directly with the senior engineer who designed, built, and runs Grounded — no juniors, no agency overhead. Let’s scope your AI support assistant or RAG system.
Also see: GrowthForge — AI Writing Assistant