AI chatbot Selangor: 5 questions to ask before a WhatsApp bot build

A 30-staff Shah Alam accounting firm, a Puchong dental chain, a Subang F&B ghost-kitchen, a Klang logistics dispatcher — Selangor's small businesses run on WhatsApp. Every owner in Petaling Jaya, Kajang, or Cyberjaya we sit down with asks the same thing in 2026: should we build a WhatsApp chatbot for our Selangor business, or is that overkill? We've shipped 14 in the Klang Valley in the last 18 months. Half paid back in 90 days, half we'd talk the owner out of if they asked us today. Here are the 5 questions that separate the two — ask these before you sign anything.

What this is and why it matters

Selangor is Malaysia's SME heavyweight — over 400,000 SMEs across Shah Alam's manufacturing belt, PJ's professional services corridor, Klang's port-adjacent logistics, Subang's aerospace cluster, Cyberjaya's tech firms, and Kajang's F&B and education scene. Most of these businesses handle customer enquiries on WhatsApp Business — the owner, a receptionist, or a sales exec typing the same answer 30-60 times a day. The honest Selangor build runs RM 8k-30k over 2-6 weeks. But a bot built on the wrong assumptions burns RM 15k-25k and the team never trusts it. The state-level case set sits on the Selangor AI agency hub; the country framing is on the AI agency Malaysia head page.

The 5 questions to ask before you build a WhatsApp bot in Selangor

1. How many WhatsApp conversations a day, and what % are the same 5-6 questions? A Shah Alam dental clinic taking 80 messages a day, a PJ property agent taking 120, a Subang tuition centre taking 45 — if 60-70% of those are the same handful of questions (price, hours, location, availability, what to bring, how to pay), a bot pays back. If the mix is complaints, custom quotes, or one-off conversations, the bot makes things slower. Count for 7 days before you build. Build only if the biggest bucket is 30+ messages a day of repeated questions.

2. What languages do your customers actually speak? Bahasa Malaysia is the default in Klang, Shah Alam, and Kajang. But Puchong and Petaling Jaya skew Mandarin-heavy; Subang and Cyberjaya skew English; Klang's Tamil estate businesses deal in Tamil. A bot that only handles Bahasa fails the moment a Chinese-Malaysian customer writes in Mandarin. Plan for at least 2 languages from day one — Bahasa + English is the floor, Bahasa + Mandarin is more common in Selangor than the rest of the country. Building a 4-language bot is a RM 5k-12k jump on top of the base. The full Selangor WhatsApp chatbot build approach sits at the state service page.

3. What existing system does the bot need to talk to? A WhatsApp bot in 2026 that doesn't connect to your booking, inventory, or CRM system is just a fancy autoresponder. Ask the vendor: can it write to SQLAccount / Xero / iBiz? Can it pull availability from your booking system (Salon Booking, SimplyBook, SiteMinder, ResDiary)? Can it check stock from your POS (StoreHub, Foodie Market, Qashier)? A bot that takes a deposit via Billplz / ToyyibPay closes the loop. A bot that just answers questions and tells the customer "please call us to book" is a waste of RM 10k. Get the integration list in writing before signing.

4. What does "escalate to a human" look like — who, when, and with what context? A bot that loops forever is worse than no bot. Define upfront: after how many failed attempts does it hand off? Who gets the WhatsApp notification? Does the handoff include the full conversation transcript, or does the staff member start cold? In every Selangor build we ship, the "human-in-the-loop" handoff is the most-iterated feature. Get this right on day one — it's the difference between a bot the team trusts and a bot they switch off at week three. RM 1k-3k of the build budget should be reserved for the handoff UX alone.

5. What does the full 12-month cost look like, and what's the payback math? Build cost is RM 8k-30k for a Selangor SME. Add the WhatsApp Business API via a BSP — Respond.io, Wati, SleekFlow, AiSensy — at RM 500-2,000 a month. Add hosting, monitoring, and a small retainer for menu and FAQ updates — RM 800-1,500 a month. A realistic 12-month total is RM 22k-58k all-in. Work backwards: if your bot saves 4 hours of staff time a day at RM 15/hr fully loaded, that's RM 1,800 a month in saved wages — payback at month 13. If it saves 8 hours, payback at month 7. Honest builds that ship at the high end have a 60-90 day payback.

Where it goes wrong

Three failure modes we see in Selangor WhatsApp bot builds. First, the bot is built without counting first — owner signs a RM 15k build based on a vendor's slide deck, then realises only 20% of messages are repeatable, the bot answers 12 questions a day, and the team keeps doing 80% of the work manually. Fix: count for 7 days, build only if the math works. Second, the bot is built before the team is trained — the front desk doesn't know how to take over from the bot, gets a weird handoff, and tells customers "ignore the bot, just WhatsApp me direct". Fix: 2 hours of staff training on launch day, plus a written 1-page handoff protocol. Third, the vendor doesn't speak Bahasa at the kickoff meeting — every prompt-engineering session needs a native BM speaker on your side, or the bot ships with broken Malay that customers screenshot and post in Facebook groups. Fix: insist the kickoff, the QA, and the month-1 review are run in Bahasa Malaysia.

What to do next

Take 7 days. Count every WhatsApp message that lands on the business number. Bucket them: order, booking, catalogue, general, complaint. If the biggest bucket is 30+ messages a day and the questions repeat daily, a bot will pay back inside 90 days in the RM 8k-30k range. Walk through the 5 questions above with your ops lead and your tech-savvy friend — if you can't answer 4 of the 5 confidently, the build isn't ready. The Selangor WhatsApp chatbot build overview walks through the 5 questions in detail and shows what a paid-back Selangor build actually looks like.