whatsapp chatbot for business in india: rule-based or ai, flows, real estate, cod and human handover

a whatsapp chatbot earns its keep in four jobs: qualifying leads, confirming cod orders, booking appointments and handing warm chats to a person. here is how to choose rule-based or ai, when to use flows, and the scripts we start from.

• this is the bot layer of our whatsapp business api guide for india.
• since 15 january 2026, whatsapp's business terms bar general-purpose ai assistants, but bots that serve a company's own customers, such as support, booking and sales bots, remain allowed (techcrunch).
• whatsapp flows are native forms inside the chat, with screens, dropdowns and date pickers, built for lead capture and booking (whatsapp business blog).
• a cod confirmation tied to a specific order can go as a utility template at about ₹0.115 before 18% gst. add an incentive and it becomes marketing at about ₹0.8631.
• for sports for life, our whatsapp ai agent delivered 5x faster first response and 200+ trial bookings in its first month.

rates live in our pricing guide and broadcasts in our whatsapp marketing guide. here we cover the bot: what to automate, how to script it and when it should step aside.

what should a whatsapp chatbot do for an indian business?

a whatsapp chatbot should answer every new chat in seconds, collect the four to six facts your team needs, take one action such as booking a visit or confirming an order, and pass anything outside its rules to a person with the full transcript. a bot that tries to answer everything ends up answering badly.

most top-ranking whatsapp chatbot guides for india come from inbox vendors listing twenty use cases. in our accounts, four jobs pay for the build. a whatsapp retention flow helped a food-tech client cut churn by 22%, using reminders and replies, not offers. pick one job and define "done":

rule-based vs ai whatsapp chatbot: which should you build?

build rule-based when answers are fixed and the customer can tap a button: prices, slots, order status. add ai when people type freely, in hinglish or regional languages, and the bot must pull facts out of messy sentences. most good bots are hybrids: buttons for the path, ai for the text that falls off it.

the sports for life agent is a hybrid. manychat runs the buttons and templates, gpt-4o reads free text and extracts sport, age and area, and n8n applies the rules (full build). the model answers only from an approved facts sheet.

our rule: start rule-based, read 200 transcripts, add ai where buttons failed.

what are whatsapp flows, and when should you use them?

whatsapp flows are native, multi-screen forms that open inside the chat, with text fields, dropdowns, radio buttons and date pickers. use them when you need clean, structured data, like a booking slot or a five-field lead form. use plain chat when the customer needs to ask questions first.

meta describes flows as a way to book appointments, browse products, collect feedback and capture leads (whatsapp business blog). meta's flows reference covers the flow json, the components and the endpoint. as 360dialog documents, a flow can be sent from a template or an interactive message (360dialog flows docs), and a business-hosted endpoint can return live data, such as free slots, while the form is open.

• use a flow for: site-visit booking, fixed-field lead forms, cod address changes, post-visit feedback.
• use chat for: "is the trial free?", "which centre is nearest?", anything asked before trust exists.
• static or dynamic: a static flow collects answers; a dynamic flow calls your endpoint, which real calendar availability needs.

we open with chat, answer a question or two, then send the flow when the customer is ready to commit.

how does a whatsapp chatbot for real estate qualify leads?

a whatsapp chatbot for real estate should qualify in five steps: configuration, budget, location, timeline and a site-visit slot. buttons keep answers clean, and every step writes to the crm. leads with budget and timeline in range go to a sales person within minutes; the rest get a slower nurture sequence.

one real estate developer client of ours averages 18 days time-to-lead, so the bot's job is to capture and sort, not close. digital can carry the whole sale: lodha sold 45 flats, 100% attributed to digital, from campaigns rather than a bot.

a typed question goes to ai if it is about the project, and to a person if it is about negotiation, loans or legal title. keep these questions inside the free window: meta's pricing documentation describes 72 free hours for chats from click-to-whatsapp ads, and respond.io reports an extension to 7 days. check meta's pricing page before planning around it.

how do whatsapp cod confirmation and abandoned cart flows work?

a whatsapp cod confirmation asks the buyer to confirm, change or cancel an order before you ship it, so fake and impulse orders are stopped before they cost two-way freight. send it as a utility template with the order number and no offer. abandoned cart reminders are marketing, priced about seven times higher.

shiprocket's own blog puts cod return-to-origin rates at roughly 20 to 30% (shiprocket, vendor-reported). every returned parcel pays shipping twice and earns nothing.

meta assigns every template to marketing, utility or authentication (meta template docs). as reported by chatmitra, offering online payment as a plain option can stay utility, but "pay online and save ₹50" adds promotional intent and makes it marketing. authentication templates are for one-time passwords only.

the flow we start from:

• minute 0: utility template: "hi {{1}}, please confirm your cod order {{2}} for ₹{{3}}, delivering to {{4}}." buttons: confirm, change address, cancel.
• confirm: mark the order ready to ship.
• change address: open a short flow for the new address, then confirm.
• no reply after 12 hours: one utility reminder, then hold the order and flag it for a call.
• cancel: cancel before dispatch and log the reason.

if it stops 300 doomed orders at an assumed ₹120 lost per rto parcel, it saves about ₹36,000. these are assumptions, not client data.

for abandoned carts, send one marketing template with an opt-out within a few hours, only to buyers who opted in.

how does whatsapp appointment booking work?

whatsapp appointment booking works in three moves: the bot answers the questions people ask before committing, a flow or button list offers real free slots, and a utility template confirms and reminds. missed appointments need their own rescue step, because a reminder alone won't bring everyone back.

this is the sports for life pattern: 60%+ of conversations qualified, 0 manual first-touch and 200+ trial bookings in month one. the case study has the build. what carries over to clinics, salons, coaching and test drives:

• slots must be real. a bot that books a full slot loses trust faster than a slow human.
• ask only what the booking needs. for a trial: sport, age group, centre and time.
• confirm and remind as utility. a confirmation and one reminder before the slot, no upsell in either.
• rescue no-shows. sports for life uses a ringg ai voice call to offer a new slot. ai voice calls in india carry their own trai rules, so check those before copying it.

whatsapp chatbot human handover: when should the bot step aside?

a whatsapp chatbot should hand over when the customer asks for a person, when it fails to understand twice, when money goes beyond published prices, when the tone turns unhappy, and when a high-intent lead is ready to buy. the person must see the full transcript, so the customer never repeats themselves.

as tyntec's help centre describes, whatsapp expects businesses that automate replies to offer a clear path to a human, such as transfer in chat, a phone number or email, and can act on quality ratings when they don't (tyntec).

what does whatsapp's 2026 ai policy allow, and what about opt-in and dpdp?

whatsapp now bars general-purpose ai assistants from its business platform, but allows ai that serves your own customers, such as support, booking and sales bots. you still need opt-in before messaging people, and india's dpdp rules require purpose-specific consent from 13 may 2027. this is general information, not legal advice.

meta added an ai providers clause to the whatsapp business solution terms, enforced from 15 january 2026, aimed at companies whose main product is the ai assistant itself (techcrunch). as respond.io explains, a bot that answers questions about your products, bookings or orders is still allowed; an open "ask me anything" assistant is not.

• scope the bot: answer from your facts sheet; refuse off-topic requests.
• opt-in: whatsapp's business messaging policy requires opt-in before you message people. a customer who starts the chat can be answered; later templates need consent.
• dpdp: the rules notified via pib require consent specific to each purpose, so separate order updates from offers.

build or buy: which whatsapp chatbot tools fit your team?

buy a no-code chatbot builder from a whatsapp provider if your flows are button-based and one inbox is enough. build on an orchestration tool such as n8n with a model and your crm when you need free-text understanding, live booking and two-way crm sync. whichever you choose, own the whatsapp account.

sports for life runs the third option. for automation beyond whatsapp, see our guide to ai automation for business in india.

whatsapp chatbot launch checklist

launch a whatsapp chatbot on a slice of traffic, with templates approved in the right categories, handover rules tested, and every test chat landing in the crm with all fields filled. read transcripts daily for two weeks before you scale the ads feeding it.

• job: one job, with "done" defined as fields.
• facts sheet: prices, timings, locations and policies the bot may quote, owned by a named person.
• templates: confirmation and reminder as utility; any offer as marketing with an opt-out.
• flows: only where structured data is needed, tested on android and iphone.
• handover: every trigger in the table above tested with a real phone.
• crm: match on phone number, store source ad, answers, consent and stage (field mapping and username changes).
• consent: wording, timestamp and source stored for every opt-in.
• metrics: first response time, qualification rate, bookings or confirmations, and handover response time, reviewed weekly.

who wrote this, how and why

who: puneetraj chavan, founder of rapchik media, who has spent 8+ years running performance marketing and automation for 30+ brands across india and the us. how: policies, template rules and rates come from meta's documentation and the sources linked above, checked on 20 october 2026. platform-reported and vendor-reported figures, such as shiprocket's rto range, were not independently audited. why: most chatbot guides are written by the tools selling bots; we wrote this because we answer for what the bot books, not how many chats it handles.

• meta and whatsapp: flows reference · whatsapp flows 101 · template fundamentals · pricing · business messaging policy
• ai policy: techcrunch on the chatbot ban · respond.io on what is allowed · uni on meta's business agent
• operations and regulation: shiprocket on rto · chatmitra on order confirmations · tyntec on human escalation · pib on dpdp rules

want a whatsapp chatbot that books, confirms and hands over, reported on outcomes? see our services or start a project.

found this useful? share it

written by puneetraj chavan

founder of rapchik media. 8+ years scaling 30+ brands across india and the us, from agri-drones to a us job portal. strategy and execution, same brain.

more about me →

want this on your brand?talk to puneetraj on whatsapp reply within 24h →