Every AI reply, with its receipts.
See exactly which FAQs and documents your AI pulled from before it replied — ranked, scored, and one click away. The black box just got a glass wall.
When the AI gets it wrong, you should know why.
Every “wait, where did it get that?” moment costs you a customer and a chunk of your team’s afternoon.
Three steps. Zero setup.
Customer asks
A WhatsApp message lands. The AI agent picks it up.
AI retrieves & ranks
It searches your Knowledge Hub, scores every match by similarity, picks the best ones.
Reply shows its receipts
The AI sends the answer with a sources pill underneath. One tap reveals the ranked list.
Everything you need to trust the answer.
Built for the question your team actually has: not “is it on or off,” but “is it right.”
Similarity scores, not vibes
Every cited source comes with a cosine similarity score (0.00 to 1.00). A 0.94 match is a confident pull. A 0.61 is a warning to check the doc. The number lives next to the filename — no guessing how strongly the AI matched.
See the chunk it actually read
Expand any source and the exact passage the model retrieved shows up — the 200 characters that made it into the prompt. Stale wording shows up before it becomes a customer complaint.
Cited vs. also retrieved
Two lists, never mixed. “Cited” is what the model actually used. “Also retrieved” is what it considered but didn’t cite. Spot the near-miss that should have been the answer — and add it to the Knowledge Hub.
Replied without sources? You’ll see that too.
When the AI answers from general knowledge — not your docs — the conversation flags it clearly. Never hidden, never implied. So you know exactly when to add an FAQ for next time.
Teams who can’t afford a wrong answer.
Education & Enrichment
When parents ask about fees, schedules, or trial classes, see exactly which programme doc the AI quoted — before a wrong number costs a sign-up.
Clinics & Healthcare
Audit every reply involving treatment info, pricing, or insurance. Compliance gets a clean trail; operators catch outdated rate cards before patients do.
Real Estate & Property
Listings change weekly. Source transparency tells you instantly when the AI is citing a unit that’s already sold — before you lose deposit-ready leads.
Professional Services
Quote ranges, scope inclusions, terms — high-stakes details where “the AI said so” isn’t an answer. Every reply ships with its receipts.
Questions you’re probably asking.
What teams actually want to know before turning on AI replies.
You'll see that too — clearly labelled, never hidden. When the model answers from general knowledge instead of your docs, the conversation flags it. That's your cue to add an FAQ or document, so next time the answer is sourced.
Still have a question?
Ask us on WhatsApp →Trust your AI.
Verify it too.
See it run on your own conversations. We’ll wire it up to your WhatsApp, train it on your docs, and walk you through the first week of source trails — free.
Trusted by Singapore’s leading teams
Explore related features