Customer · Operations · MODULE A
WhatsApp AI Receptionist
Your WhatsApp gets 200 messages a day. Most don't need you.
Module A picks up every WhatsApp message in BM, English, Mandarin, or Manglish — answers from your knowledge base, books appointments, qualifies leads, detects anger BEFORE complaints escalate, and only hands off to your team when it genuinely needs a human (with a 3-line summary + suggested first reply attached). Built for Malaysian SMEs, not retrofitted from a US tool.
24/7
Always-on coverage
4 languages
BM · EN · 中文 · Manglish — code-switch mid-turn
60-70%
Inbound deflection (industry-leading)
WHAT THIS FIXES
The exact problems WhatsApp AI Receptionist is built around.
- You're losing customers because WhatsApp messages outside business hours go unanswered.
- Your team is buried in repeat FAQs ("Do you open Sunday?", "What's the address?") instead of doing real work.
- Foreign chatbots butcher Manglish and BM — customers detect the script in 2 messages and bounce.
- Hiring a real receptionist costs RM 2,500-4,000/mo and they still can't handle BM + Mandarin natively.
- When complaints DO come in, you find out 6 hours later — by then the customer is already on Google Reviews.
HOW IT WORKS
The pipeline, end to end.
- 1
Knowledge base ingested
Upload your FAQ doc, website, PDFs, or paste text. Module A indexes everything with embeddings for semantic retrieval (RAG-grounded — no hallucinations).
- 2
WhatsApp Business connected
Plug your existing WhatsApp Business number into our official ASP — Module A becomes the agent on every inbound.
- 3
Conversation handling
Every message: language-detected → intent-classified → sentiment-scored → KB-searched → reply generated in customer's language → reply within 8 seconds.
- 4
Sentiment-driven auto-escalation
If sentiment drops below 0.30 (angry), Module A escalates IMMEDIATELY — even before the LLM decides. Your team gets the alert before the complaint amplifies.
- 5
Handoff with context
Every escalation includes a 3-line conversation summary + a suggested first reply for your human agent + the customer's sentiment trend + intent label.
- 6
Topic trends + KB gap digest
Weekly digest: top customer intents, sentiment distribution, top unanswered questions clustered into KB gap suggestions. Self-resolve % goes up every week.
WHAT'S INCLUDED
Every capability shipped, named.
- BM, English, Mandarin, and Manglish — native fluency with mid-turn code-switching
- Knowledge-base-grounded RAG (vector search, no hallucinations)
- Sentiment scoring on every message + auto-escalation below 0.30
- Intent classification (sales / support / complaint / booking / FAQ / feedback / spam) — feeds analytics
- Working-hours awareness — different tone + callback promise outside business hours
- Handoff with auto-summary + suggested first reply for the human agent
- CSAT collection via WhatsApp quick reply after every resolved chat
- Knowledge-gap weekly digest — clusters unanswered questions for owner review
- Appointment booking via Module Q (calendar integration)
- Lead scoring + Module N hand-off for sales pipeline
- PDPA-native: consent log, purpose tagging, 90-day default retention, 1-click DSR
- Prompt-injection guard + rate limit + LLM budget cap (silent runaway protection)
- Topic trends dashboard — see what customers actually ask about week-over-week
- Multi-tenant — agencies can deploy Module A per client
- Works with WhatsApp Business or WhatsApp Cloud API
PRICING
Plans that scale with use, not seats.
All prices in MYR, exclusive of SST. 14-day refund on monthly, pro-rated refund on annual. PDPA-aligned. Cancel anytime.
WHAT YOU RECEIVE
The artefacts that land in your hands.
Inbound response in <8s
Customer sends WhatsApp → Module A receives via webhook → retrieves KB → generates response → sends back. End-to-end median 6.2 seconds. Every message also gets sentiment-scored, intent-classified, and language-detected — all in the same single LLM call (zero added latency).
Topic trends + KB gap weekly digest
Every Monday: top intents (sales/support/complaint/booking/FAQ), sentiment distribution, top unanswered questions clustered for KB additions, handoff rate, CSAT average. Lift self-resolve % each week.
Escalation with full handoff package
When sentiment drops below 0.30, customer asks for human, or LLM flags complex/legal/payment-dispute — Module A pauses + sends your team a Telegram alert containing: 3-line conversation summary, sentiment trend, intent label, suggested first reply for the human agent. Median handover time: 12 seconds.
CSAT collection in-chat
After every resolved conversation, Module A asks "Rate this conversation 1-5 stars" via WhatsApp quick-reply. Scores feed the analytics dashboard + flag drops for owner review.
FREQUENTLY ASKED
The questions buyers ask first.
Does it sound like a robot?
No — Module A is trained on natural Malaysian conversational patterns (kahwin, makan, ok lah). Customers usually don't realise they're talking to AI until you tell them. We can also force a "AI assistant" prefix if you want full transparency.How is this different from Wati / AiSensy / Respond.io?
Three real differences. (1) Native Manglish — those tools rely on user-written templates per language; Module A code-switches mid-conversation because DeepSeek + our prompt handles it natively. (2) Sentiment-driven auto-escalation — we escalate angry customers BEFORE the AI decides, on score ≤ 0.30. Most competitors only escalate when the customer types "speak to manager". (3) PDPA-native — purpose tagging, 1-click DSR, MY data residency, consent-log audit trail. None of the foreign tools have this.How does it handle my prices / SKUs / specifics?
Whatever you put in your knowledge base. We support PDFs, Notion, Drive, Confluence, plain text. KB updates take effect within 60 seconds. The chatbot uses vector search to find the most relevant chunks for every question — no hallucinations.What about WhatsApp Business approval?
We integrate with WhatsApp Business and WhatsApp Cloud API. If you don't have a Cloud API approval yet, we provide a setup guide (~2 days end-to-end).Can it process payments?
Not in conversation — that's a Mandate 2.4 cancel-window violation. It can send a Stripe / Billplz link generated by Module Z, customer pays externally, payment confirmation flows back via webhook.PDPA-compliant?
Yes — first-in-Malaysia. All messages encrypted at rest, retention configurable per tenant (default 90 days), purpose-of-processing tagged on every conversation, consent-log on every contact. Data subject access requests are 1-click via /admin. No foreign chatbot offers this.What about handover to humans?
Three triggers: (1) sentiment_low_water ≤ 0.30 (angry customer), (2) explicit "speak to human" intent detected by the LLM, (3) LLM-flagged complex topic (legal/payment dispute). Each handoff includes a 3-line summary + suggested first reply for your agent. Median handover time: 12 seconds.Can I see what customers are asking about?
Yes — the /admin/chat-analytics dashboard shows: total conversations 7d, self-resolve %, handoff %, avg sentiment, avg CSAT, this week's topic distribution, top knowledge gaps (clustered questions the bot couldn't answer), and recent escalations with full context. Updated live.
Try WhatsApp AI Receptionist on your real business.
14-day trial. No credit card. PDPA-aligned. Built for Malaysian SMEs by a Malaysian team in KL. Reginald and ONSET team reply within one business hour, Mon–Fri 9am–6pm MYT.
- PDPA-aligned
- Cancel anytime
- 14-day refund
- 1 business hour reply
SEE OTHER AI EMPLOYEES
The other 24 modules in the platform.
Each AI Employee is independently usable, but they compose. Run a demo and we'll show you which combination fits your business.