From choosing to shipping
Use product information and delivery policies as context.
AI Customer SupportPilot · assessment required
Answers for everyday questions.More room for human care.
Turn company information into answers customers can find on your website and LINE, with sources and a path to human help.
From a question to an answer with sources.
Another repeat question? See how your AI assistant answers from your information and hands over to your team when needed.
● ● ●AI support · example conversation
Illustrative example · Answers depend on the information provided
A different way to handle the questions that keep coming.
Let your AI assistant answer common questions, so your team can focus on the care that needs a human touch
Stuck? Find your next step
Help customers find their next step in guides and FAQs.
Human help when it matters
Ask about something the material does not have and it says the data is unconfirmed and gives your contact instead of inventing an answer; unanswered questions notify the team.
Illustrative use cases. Answers depend on the information provided.
Support where your customers already are.
The website widget and LINE share one knowledge base, so your team maintains company information in one place.
Keep browsing. Keep talking.
Ask while browsing products and keep the conversation in a compact window.
One knowledge base, now in LINE
Our own LINE account uses this assistant. Ask a question and inspect the answer and sources.
Illustrative use cases. Answers depend on the information provided.
Grounded in your information. Guided by your team.
Review imported documents, answers, and unanswered questions to keep improving your knowledge base.
Explore setup and delivery ↗01
Document refresh: hourly knowledge updates
02
Human handoff: notify the team when no answer is available
03
Weekly report: review gaps in the knowledge base
ON YOUR HARDWARECompany knowledge
Company knowledge
on your own hardware
On-premise installation is available for the assistant and index. Review your use case, hardware and maintenance requirements before choosing a deployment.
Explore on-premise deployment ↗Optional deployment, subject to use case and hardware review▣ YOUR COMPANYDeployment illustration
▤Your documentsPDF · Word · Excel
✦Local AI & indexFind relevant context
A TEAM QUESTIONDoes this document have the answer?
↳
Search company information and check the original source
Hi, I’m Navi
MEET NAVI
A question? Navi is here
Wondering how AI could help? Start with what you need
Chat in the corner and keep exploring
Explore testing and technical details
The assistant on this site, one real run
Recorded September 24, 2026; the counts are from that run
Can In-house Data Lookup run with a local model? Does our data leave the building?
AskCan In-house Data Lookup run with a local model? Does our data leave the building?
Yes, In-house Data Lookup can run with a local model 1. Your data does not leave the building, as documents, indexing, and selection all occur in your environment 3.
How it answered
- Retrieve4/1802The site is cut into pieces and every one is scored against your question. Only the highest scoring few are put in the prompt; the model never sees the rest.
- Answer2/4The model may answer from those pieces and nothing else. What it knows from anywhere else does not count here.
- Verify2/2Every citation it wrote is matched back against what was actually sent, and a number or an address in a cited sentence has to appear on that page. A marker that fails either check is left as plain text instead of becoming a link.
- Answered from
- [1]In-house Data Lookup100%
- [2]Do you do custom or on premise work?99%
- [3]Put the knowledge in the index, and the model can stay in the building › Why this pairs so well with a local model（1/4）94%
- [4]Put the knowledge in the index, and the model can stay in the building › We built it as In-house Data Lookup（2/2）90%
Every number in the answer links to the page that sentence came from.
What happens when the harness comes off.
Same model, same 28-passage corpus, same 30 questions. The only thing removed is the code around the model: picking the passages, stopping when none of them match, and checking the citations afterwards.
| Measure | With the harness | Without it |
|---|---|---|
| Passages sent to the model | 3 (240 characters) | 28 (2,076 characters) |
| Input tokens per question | 497 | 2,768 |
| Cost per question, one assumed rate card | $0.000647 | $0.002959 |
| Questions answered acceptably | 30 / 30 | 23 / 30 |
| When the answer is not in the documents | The model is never called | A fluent answer, with a citation that does not match it |
| Claimed a source that could not be verified | 0% | 20% |
Two runs, both 2026-09-02 on our own hardware. Passage and character counts come from one pass (30 questions × 3); tokens, cost and the verification rates from the other (30 questions × 1). The rate card is an assumption used to show the arithmetic, not a quote. Twenty-eight passages is a small corpus and none of this transfers to your documents until it has been re-run on them.
YOUR NEXT STEPTurn everyday questions
Turn everyday questions
into your next step
Start with one common question. Together, we can define what belongs with AI and what needs your team.
A SIMPLE START
01
What do customers ask?
Choose one recurring question
02
Where is the answer?
An FAQ, service guide or product page
03
No documents ready? Tell us what you have in mindWhere should support happen?
Your website, LINE, or a use case to discuss