Concept buildA worked example of how we would build this for a typical business. Not client work: no client, no client data, no achieved results.
AI Chatbots
Concept buildA concept build for a direct-to-consumer brand in India whose WhatsApp inbox fills with the same two questions: where is my order, and how do I return it. The assistant answers in English, Hindi or Hinglish from live order data and the brand’s own returns policy, and hands anything unusual to a person with the whole thread attached.
The brief
A typical D2C brand selling through its own online store runs most customer conversations on WhatsApp, because that is where its buyers already are. In a sale or festive season the inbox turns into a queue of near-identical messages — “order kab aayega?”, “COD order cancel karna hai”, “size exchange ho sakta hai?” — answered by a small team pasting tracking links by hand. Replies slip from minutes to hours, customers ask twice, and the hard conversations, such as a damaged parcel or a refund that has not arrived, wait behind the easy ones.
Explainer
Explainer · 7 slides · about 45 s
Concept buildConcept build: Hinglish WhatsApp Order Assistant for a D2C Brand
Concept build, explained in 7 slides: a concept build for a direct-to-consumer brand in India whose WhatsApp inbox fills with the same two questions: where is my order, and how do I return it.
Slide 1 of 7
01 · AI chatbots
Hinglish WhatsApp Order Assistant for a D2C Brand
Concept build: Hinglish WhatsApp Order Assistant for a D2C Brand.
02 · The problem
What a typical D2C brand is up against
A typical D2C brand selling through its own online store runs most customer conversations on WhatsApp.
03 · The approach
How we would build it, in 5 steps
We would build it in 5 steps, from intent audit on a 60-day export to handoff inbox and a measured pilot.
04 · Architecture
How the pieces connect
How the parts connect, from the first input to the person who reviews the result.
05 · Demo · sample run
A Hinglish order question, a return and a handoff
A sample run, step by step, including where a person steps in.
06 · Typical targets
What we would measure it against
Typical targets, each with the basis it would be measured against.
07 · Next step
Scope a build like this for your business
Book a strategy call to scope a build like this.
Architecture
Architecture of the Hinglish WhatsApp Order Assistant for a D2C Brand concept: WhatsApp → LLM + retrieval (message); Returns policy → LLM + retrieval (cited clause); Store order API → LLM + retrieval (order status); Courier tracking → LLM + retrieval (tracking); LLM + retrieval → Team inbox (handoff); LLM + retrieval → Return requests (new request).
Diagram with 7 parts: WhatsApp (source), Returns policy (source), Store order API (processing), Courier tracking (processing), LLM + retrieval (model), Team inbox (human), Return requests (system of record).
Connections:
Sample run
Customer: “order kab aayega?”Routine
The assistant asks for the order ID or the phone number used at checkout.
Order verified and looked upRoutine
Status: shipped. Tracking link and expected delivery date sent in Hinglish.
Customer asks about returnsRoutine
A policy answer with the clause cited: the window and the condition rules.
Return request loggedRoutine
The request goes to the team queue; refund approval stays with a person.
Customer: “talk to a person”A person takes over
The thread moves to the team inbox with transcript, order and intent.
Build plan
We export about sixty days of WhatsApp and email threads and label them by intent: order status, cancellation, return or exchange, payment, product questions, complaints. The audit shows which intents have one clear source of truth, either the order system or the written policy, and those become the launch scope. Everything else stays with people.
The assistant reads order status, courier tracking and return eligibility through the store’s order API, and only after the customer confirms the order ID or the phone number the order was placed with. It sees one order at a time. It can log a return or exchange request, but approving a refund always stays a human decision.
Returns, exchange and cancellation rules go into a short policy knowledge base indexed for retrieval. Each policy answer quotes the clause it came from. When the policy does not cover a question, the assistant says so and offers a person instead of improvising an answer.
Replies follow the language the customer writes in, whether English, Hindi or Hinglish, and stay inside WhatsApp Business Platform rules: free-form replies only within the customer-service window, pre-approved templates for anything the brand sends first, and an opt-out honoured at once.
“Talk to a person”, an angry tone, a payment dispute or a low-confidence answer moves the thread to the team inbox with the transcript, the order and the detected intent attached. The pilot starts on a share of conversations; we compare resolution, handoff rate and repeat contacts per intent before widening it.
Typical targets
Typical target
First reply on web chat and WhatsApp in under a minute, any hour
Basis: Set by the automation rather than by staffing, on the questions in the launch scope. Measured in the pilot from message received to first reply.
Typical target
50–70% of order-status questions answered without an agent
Basis: “Where is my order” has a deterministic answer in your order and carrier data; the range allows for unverified customers, stale tracking and complaints. Confirmed from a sample of your recent tickets before launch.
Typical target
Every policy answer shows the clause it came from
Basis: Design property: an answer without a retrieved policy passage is not sent
Typical target
No refund approved by the assistant
Basis: Design property: refund decisions route to a person and are never executed by the model
Your project
Questions taken from your tickets and transcripts, each paired with the source that answers it, with a pass threshold agreed before launch.
The knowledge index, model accounts and channel numbers are set up under your business, so the data and the configuration stay yours.
Message templates, consent wording and widget placement checked against each channel’s rules where your customers are.
Handoff and resolution tracked per intent; scope widens only when the numbers hold for a full week.
Where this fits
FAQ
No. No brand runs this assistant: it is a concept build showing how we would design WhatsApp order support for a typical Indian D2C brand. The Hinglish conversation in the demo is sample data, and the resolution range is a typical target tied to an intent audit, not a result measured for a client.
That is the normal case it is designed for. It replies in the mix the customer uses, and the pilot test set includes Hinglish phrasings taken from the intent audit, so misread intents show up before launch. Other Indian languages are added only after they pass the same test.
The lookup runs only after the customer confirms details that match the order, and the assistant’s API key can read one order at a time. It has no access to payment data and cannot change an order’s payment or refund status.
Only through message templates the brand has had approved for WhatsApp Business messaging, and only to customers who opted in. A delay notice is a template triggered by the courier status, not a free-form message. Marketing broadcasts are outside this design; they need their own consent and a separate review.
The assistant answers in-scope intents in parallel, so the part of the queue it removes is the repetitive part. The handoff rules do not change under load: complaints, payment disputes and angry messages still reach people first. We run the pilot before a sale rather than during one, so thresholds are tuned on normal traffic.
Worked examples
Concept builds are worked examples, not client work.
Concept buildAI Chatbots
A concept build showing how we would build a support bot for an online electronics store whose helpdesk fills with the same few questions. The bot would answer order, delivery, returns and product-fit questions from the store’s catalogue, policies and live order data, on the website and WhatsApp, and pass everything else to the helpdesk with the conversation attached.
Concept buildAI Chatbots
A concept build for a B2B software company whose help centre is thorough but hard to search. The copilot answers product questions from the published documentation, shows the passages each answer came from, and opens a support ticket, carrying the question and the sources it tried, whenever it is not confident.
Concept buildAI Chatbots
A concept build for a boutique hotel in the UAE whose front desk answers the same guest questions in several languages, day and night. The concierge replies on WhatsApp and the hotel website in the guest’s language, answers from the hotel’s own guide, turns requests into housekeeping tickets with the room number, and puts complaints in front of the duty manager straight away.
Next step
Book a free 30-minute strategy call. Bring one process you want to automate and we will talk through scope, integrations and a sensible first step.
Free 30-minute call · no commitment
