Replies in under 10 seconds
Drafted automatically the moment a lead messages, even at midnight. You stay in approval mode for the first month.
All features
AI agent · Aurora
An AI agent that replies in your voice.
Train Aurora on your business — pricing, FAQs, qualification rules, what to never say. She drafts replies in seconds, hands off to your team when the deal needs human judgement, and never invents a fact.
Approval-first by defaultAccount-level trainingNo hallucinations
Drafted in 4 seconds
app.botsync.in/inbox — AI Aurora
Hi, what's the fee for the JEE 2-year programme? And do you have hostel?
Priya · 11:02
Aurora drafted a reply
3 sourcesHi Priya! Our JEE 2-year programme is ₹1,80,000/year (₹3.6L total). Yes, we have separate AC and non-AC hostel options on campus — ₹85,000/year for AC, ₹55,000/year non-AC. Would you like me to share the brochure and connect you with a counsellor?
Knowledge: JEE-2yr-pricing.pdfKnowledge: hostel-rates.pdf
Auto-handoff in 2 turnsSafe-fact guard active
Aurora has drafted 1.2 million replies across coaching, real estate, clinics, and SMB e-commerce
What makes Aurora different
Trained on your business, not the internet
Off-the-shelf chatbots hallucinate. Aurora is account-level — your prompts, your knowledge, your guardrails — so replies feel like they came from your best counsellor, not a generic LLM.
Grounded in your knowledge
Aurora answers from your uploaded PDFs, docs, and FAQs. Every claim is traceable to a source.
Knows when to hand off
Detects discount asks, escalation language, and complex objections — escalates to a human with full context.
Account-level training
Your business has its own AI — not a global one
Most platforms have a single prompt for every customer. Aurora is configured per-workspace, so the AI your real-estate client uses doesn't share knowledge with your coaching client.
- Workspace-scoped instructions (tone, style, dos and don'ts)
- Per-workspace knowledge collections — PDFs, DOCX, plain text, URLs
- Industry-specific guardrails (fees, cutoffs, placements, medical advice)
- Language preferences — Hindi, English, Hinglish, Tamil, more
Aurora training
SavedInstruction
You are a counsellor at NIET. Reply warmly, in English or Hinglish based on the student's message. Never quote fees you don't have a source for.
Tone
Warm · Concise
Language
Match user
Never say
guaranteed placement100% scholarshipJEE rank promise
Approval-first by default
Aurora drafts. You approve. Or take over.
Every workspace starts in approval mode — Aurora drafts the reply, the team reviews it, one tap to send. Switch to auto-send per-workspace once you trust the tone.
- Per-conversation Pause/Resume toggle
- Per-workspace AI-only vs Approval-first vs Off
- Edit drafts inline before sending
- Skip drafts you don't want — Aurora won't try again on that thread
Workspace mode
Approval-firstAI only
Aurora replies automatically. Use after you trust the tone.
Approval-first
Aurora drafts, you approve. Default for new workspaces.
Off
No AI on this workspace. Manual replies only.
This conversation
Knowledge with citations
Every fact comes from a source
Aurora reads your uploaded knowledge base on every reply. Every numeric claim shows the source chip — so you can verify or update the source if it's stale.
- Drop in PDFs, DOCX, TXT files — auto-chunked and embedded
- Workspace knowledge collections — separate KBs per product line
- Source citations show on every AI draft
- Safe-fact guard refuses to invent numbers it doesn't have
Knowledge collections
12 docsJEE-2yr-pricing.pdf
248 KB
hostel-rates.pdf
112 KB
scholarship-rules.pdf
87 KB
faqs-admission.docx
44 KB
syllabus-2026.pdf
1.2 MB
Every AI reply cites the document it pulled from. Stale source? Re-upload and Aurora re-indexes in under a minute.
Smart handoff
Knows the moment to bring in a human
Detects discount asks, complex objections, refund requests, legal queries, and escalation language. Hands off to the right team with a clean conversation summary.
- Trigger handoff on keywords ("manager", "refund", "complaint")
- Auto-pause AI when the lead's sentiment turns sharp
- Generates a 3-line summary for the human taking over
- Stage auto-changes when handoff happens
Auto-handoff triggered
Lead mentioned "discount" — handoff to senior counsellor
Can you give me a 20% discount? Other coaching is offering it.
Handoff to human
Aurora paused. Conversation assigned to Anand R. (senior counsellor).
Summary for Anand
Priya is interested in JEE 2-year programme. Wants discount (mentioned competitor pricing). Hostel preference: AC. Budget unspecified. Started chat 12 min ago.
Safety & control
Built for the way you sell, not the way an LLM thinks
AI replies are useful when they're bounded. Aurora is designed to refuse, defer, and hand off — not to fill gaps with invented confidence.
Safe-fact guard
Aurora refuses to quote prices, cutoffs, or guarantees she doesn't have a source for.
Workspace isolation
Your knowledge and prompts never leak to another customer's AI.
Pause anytime
Per-workspace and per-conversation kill switches. AI never speaks when you don't want it to.
Full audit trail
Every AI draft, edit, and send is logged with timestamp, model, and source citations.
From upload to auto
Ship an AI agent in a day — not a quarter
Upload knowledge
Drop your pricing sheet, FAQs, brochures into the workspace knowledge collection.
Set instructions
Write the persona, what to collect, what to never say. Per-workspace.
Approve drafts
Aurora drafts replies; your team approves with one tap for the first month.
Switch to auto
Once you trust the tone, flip to AI-only mode. Aurora hands off when needed.
Why not other AI chatbots
Aurora vs generic AI plug-ins
Most WhatsApp tools ship an AI tab that's really a thin wrapper around ChatGPT. Aurora is the opposite — built from the ground up to be bounded, accountable, and handover-aware.
| Feature | Aurora | Generic AI add-ons |
|---|---|---|
| Account-level training (not a shared prompt) | Single global prompt | |
| Knowledge base with source citations | Unattributed or none | |
| Approval-first mode | Often auto-send only | |
| Safe-fact guard (refuses invented numbers) | Hallucinates freely | |
| Per-conversation pause | Workspace-wide only | |
| Auto-handoff with conversation summary | Hand-off but no context |
FAQ
Common questions about Aurora
Will one customer's training affect another's?+
No. BotSync is architected around per-workspace configuration. Each workspace has its own prompts, knowledge base, language preferences, and guardrails. Your AI never sees another customer's data.
Can a human review every AI reply?+
Yes. Every new workspace starts in Approval-first mode. Aurora drafts the reply, your team taps Approve, edits, or skips. You can switch to AI-only once you trust the tone.
Which AI model does Aurora use?+
GPT-4o-mini is the default for chat replies. You can plug in your own OpenAI key (and other providers) at the workspace level. We never train on your data.
What happens if Aurora doesn't know the answer?+
She says she doesn't know and asks if she should connect the lead with a human. We deliberately tuned Aurora to refuse instead of hallucinating — especially for prices, cutoffs, and policies.
Can I see what Aurora used to draft a reply?+
Yes. Every AI draft shows source chips (which knowledge documents it cited) and a 'See reasoning' link that expands the system prompt + retrieved chunks used.
How long does training take?+
PDF + DOCX files index in under 60 seconds. URLs are fetched and chunked in a similar window. Aurora is ready to reply with your knowledge as soon as the indexing badge turns green.
Pair it with
What works well with Aurora
Train Aurora on your business today
14-day free trial. Upload your knowledge base, write your instructions, and watch Aurora draft replies in under 10 seconds.