WhatsApp AI Agent

A WhatsApp AI agent for your cloud kitchen: what it does and where it stops

A chatbot follows the paths you draw. An AI agent reads what the customer actually wrote, looks up your cloud kitchen's own information, and takes the next step — books, updates, follows up. This is what that looks like for cloud kitchens, and how to keep it honest.

Riya · ReputooUpdated 25 September 202610 min read

✓ Fact-checked against Meta’s WhatsApp rate card and vendors’ own pricing pages

Example WhatsApp conversation for cloud kitchens — WhatsApp AI agent
An example WhatsApp AI agent conversation for cloud kitchens. Names and details are placeholders.
Key takeaways
  1. 01The AI agent takes orders in the customer's words, answers menu and dietary questions, and tracks deliveries.
  2. 02It knows the menu with allergens and what is unavailable today.
  3. 03Complaints and refunds go to staff; severe allergies are confirmed by a person.
  4. 04It remembers regulars' usual orders and addresses.
On this page · 39 chapters
  1. 01 · What customers ask, and what the agent can answer
  2. 02 · Jobs to give the agent
  3. 03 · The knowledge base for a cloud kitchen
  4. 04 · How it learns your business
  5. 05 · Orders in natural language
  6. 06 · Where cloud kitchens should keep a person in the loop
  7. 07 · Chat and phone, one agent
  8. 08 · Recommending dishes
  9. 09 · Complaints
  10. 10 · Payments
  11. 11 · Languages
  12. 12 · The first two weeks
  13. 13 · Meal plans
  14. 14 · Bulk orders
  15. 15 · Delivery-area questions
  16. 16 · Busy-time honesty
  17. 17 · Making it sound like your brand
  18. 18 · Reading the conversations
  19. 19 · Privacy
  20. 20 · Pickup orders
  21. 21 · Cross-brand orders
  22. 22 · Hygiene and licence questions
  23. 23 · After-delivery check
  24. 24 · Loyalty for direct customers
  25. 25 · Kitchen pauses and last orders
  26. 26 · Menu changes
  27. 27 · Handovers
  28. 28 · Late-night orders
  29. 29 · Addresses and landmarks
  30. 30 · Customisation limits
  31. 31 · Office orders and invoices
  32. 32 · Tone
  33. 33 · Lunch pre-orders
  34. 34 · Group orders
  35. 35 · Payments and refunds
  36. 36 · Reorders from history
  37. 37 · Weekly review
  38. 38 · Delivery instructions
  39. 39 · Frequently asked questions
Chapter 01

What customers ask, and what the agent can answer

QuestionAgent answers?What it needs
Is the biryani spicy? Is there a veg option?YesMenu with tags
Do you deliver to my area?YesDelivery area
Where is my order?YesOrder status
Can you make it less oily?Yes, as a note to the kitchenCustomisation rules
Is it safe for a peanut allergy?No — staff confirmA person
How the AI agent handles cloud kitchen enquiries: Customer asks a question, Agent reads your knowledge base, Answers, books or quotes, Confirms on WhatsApp, Person takes over on request
How the AI agent handles cloud kitchen enquiries
Chapter 02

Jobs to give the agent

Start with what customers ask. Customers ask where their order is, whether an item is available today, whether you deliver to their area, and why the order they got differs from the one they placed. An agent answers all of that from your own information. Then give it the work cloud kitchens repeat every day:

  1. Customer orders directly on WhatsApp — Bot takes the order, generates invoice, and collects payment — zero commission.
  2. Live delivery updates sent automatically — Order confirmed, out for delivery, and delivered — each update sent on WhatsApp.
  3. Post-delivery re-order prompt — After delivery, an automated thank-you with a one-tap re-order link.
  4. Win-back for 7-day inactive customers — Customers who haven't ordered in 7 days get a personalised re-order offer.

The agent runs these on WhatsApp in the customer's language, day and night, and writes what it did back to the contact record so staff see the full history.

Chapter 03

The knowledge base for a cloud kitchen

  1. Menus per brand with prices, portions, spice levels and allergen tags.
  2. Delivery area and times, charges and minimum order.
  3. Hours and last orders per brand.
  4. Customisation rules — what the kitchen can and cannot change.
  5. Policies — refunds, missing items, cancellations.
Chapter 04

How it learns your business

An AI agent is only as accurate as what you give it. On Reputoo it answers from your knowledge base — pages crawled from your website, documents and price lists you upload, and policies you write down — rather than from general knowledge. Two rules keep it right:

  • If it is not in the knowledge base, the agent says it will check and hands over, rather than guessing.
  • Prices, timings and policies live in one place. Update the source and the agent's answers change with it.

It also needs to see the records your cloud kitchen already keeps — what to connect: aggregator dashboards plus a kitchen display or POS; direct WhatsApp orders should land in the same queue as everything else.

Chapter 05

Orders in natural language

"Two chicken biryanis, one extra spicy, a raita and a Coke, same address as last time" — the agent maps it to the menu, applies the customisation, confirms the total and time, and sends it to the kitchen.

Chapter 06

Where cloud kitchens should keep a person in the loop

These are the moments customers remember, and they are worth a human:

  • Zomato/Swiggy commissions eat your margin.
  • Customers order once, platforms own them. Zomato knows your customers. You don't. They can recommend competitors on your next order.
  • Repeat orders leak to the aggregator. A customer who ordered direct last week orders through Zomato this week, because nothing reminded them you exist.

The agent can open each of these conversations and collect the details; a person should close anything involving a complaint, an exception to policy, or money that has not been agreed. Handover goes to the shared inbox with the full conversation attached.

Chapter 07

Chat and phone, one agent

The same agent that answers on WhatsApp can answer your cloud kitchen's phone line as Vajra, Reputoo's voice agent, so a customer who calls gets the same answers as one who messages. It speaks 36 languages and switches to the caller's.

Chapter 08

Recommending dishes

"What's good for four people under ₹1,000?" — the agent suggests combos from the menu within budget. It never invents dishes or prices.

Chapter 09

Complaints

The agent apologises, collects the order and a photo, and hands to staff. It never decides refunds or credits.

Chapter 10

Payments

It sends payment links and confirms payments. Payment failures go to staff. It never asks for card numbers, PINs or OTPs.

Chapter 11

Languages

The agent replies in the customer's language across 36 languages and understands dish names as customers say them.

Chapter 12

The first two weeks

  1. Load menus with tags, delivery rules and policies.
  2. Start with regulars and reorders.
  3. Review conversations daily and add dish names it misses.
  4. Open it to all direct customers.
Chapter 13

Meal plans

The agent explains meal plans — menu, days, price, pause and skip rules — enrols customers with a payment link, and handles skips, pauses and address changes. Renewals are reminded before the plan ends.

Chapter 14

Bulk orders

For office and party orders, the agent collects the date, time, headcount, dishes and address, suggests platters from the menu, and hands to the manager to confirm the final price and timing.

Chapter 15

Delivery-area questions

The agent checks whether an address is in the delivery area and gives an honest time. Outside the area, it suggests pickup or the aggregator where you are listed.

Chapter 16

Busy-time honesty

During the dinner rush the agent gives longer, accurate times based on the kitchen's queue, and can stop taking orders if the kitchen pauses. Customers forgive a longer wait they were told about.

Chapter 17

Making it sound like your brand

Each virtual brand can have its own voice — a biryani brand warm and generous, a salad brand crisp and light. Set the agent's tone per brand number, and read conversations weekly.

Chapter 18

Reading the conversations

A month of conversations shows which dishes people ask for that you do not make, which customisations are common, and when orders peak. Those shape the menu and the staffing of each shift.

Chapter 19

Privacy

The agent shares order and address details only with the customer on record, never with another caller.

Chapter 20

Pickup orders

For customers nearby, the agent offers pickup with a ready time and directions to the pickup window, without a delivery charge.

WhatsApp Business API cost in India: marketing ₹0.8631, utility ₹0.115 per message (Meta rate card, October 1, 2026)
Meta's per-message prices in India (rate card dated October 1, 2026), before GST. Reputoo adds no markup.
Chapter 21

Cross-brand orders

When a customer asks one brand for another brand's dish from the same kitchen, the agent can take both in one order if your system allows, or share the other brand's link. Customers get what they want; the kitchen keeps the order.

Chapter 22

Hygiene and licence questions

The agent answers factually from what you have written — licence number, how vegetarian and non-vegetarian items are handled, packaging — and passes anything else to the manager.

Chapter 23

After-delivery check

After delivery the agent asks how the meal was and handles the reply — thanks for good feedback, an apology and handover with a photo request for problems.

Chapter 24

Loyalty for direct customers

The agent tells direct customers their loyalty balance, applies rewards to orders, and explains the rules. It never creates rewards outside the rules.

Chapter 25

Kitchen pauses and last orders

When the kitchen pauses orders or reaches last orders, the agent stops taking orders and says when ordering resumes, rather than accepting food the kitchen cannot make.

Chapter 27

Handovers

Complaints, refunds, severe allergy questions and bulk orders go to staff with the order and the customer's words, and the agent says when a reply will come — within minutes for food issues.

Chapter 28

Late-night orders

For late-night kitchens, the agent takes orders at any hour the kitchen is open, and after closing tells customers exactly when ordering opens again.

Chapter 29

Addresses and landmarks

The agent asks for a landmark or location pin when an address looks incomplete, saves it, and reuses it next time — fewer lost riders, hotter food.

Chapter 30

Customisation limits

The agent accepts only customisations the kitchen can make, and declines others politely before the order is placed. Customers prefer an honest no to a promise the kitchen breaks.

Chapter 31

Office orders and invoices

For office customers, the agent takes team orders, suggests packages, and sends GST invoices in the company's name, or adds the order to the monthly account.

Chapter 32

Tone

Food ordering should feel quick and friendly. Keep the agent's replies short, confirm exactly what was ordered, and save longer conversations for complaints, where the customer needs to feel heard.

Chapter 33

Lunch pre-orders

The agent takes lunch pre-orders in the morning until the cut-off, confirms the delivery time, and adds them to the kitchen's plan. Office regulars can set a recurring order and change it by message.

Chapter 34

Group orders

For group orders, the agent takes everyone's choices from the organiser — "three veg biryanis, two chicken, one without onion" — reads them back, and sends tracking to the organiser.

Chapter 35

Payments and refunds

The agent sends payment links and confirms payments. Refunds for missing items are approved by staff and then processed; the agent tells the customer when to expect them. It never asks for card numbers, PINs or OTPs.

Chapter 36

Reorders from history

"Same as Friday" is enough for the agent: it finds Friday's order, confirms items, address and total in one message, and sends it to the kitchen.

Chapter 37

Weekly review

Read the agent's handovers and misunderstood orders weekly with the kitchen lead. Add dish names, customisations and delivery areas it missed; accuracy improves every week.

Chapter 38

Delivery instructions

The agent saves instructions like "leave at the gate" or "call on arrival" with the address and adds them to the rider's ticket, so the rider does not need to call the customer or the kitchen.

Chapter 39

Frequently asked questions

What is the difference between a WhatsApp chatbot and an AI agent?

A chatbot follows menus and flows you define. An AI agent understands free-form messages, answers from your own knowledge base, and can take actions such as booking or updating a record. Most businesses run both: flows for fixed tasks, the agent for everything else.

Can an AI agent make mistakes with my cloud kitchen's customers?

Yes, which is why it should answer only from your knowledge base and hand over when it does not know. Review conversations in the first weeks and add missing answers to the knowledge base as you find them.

Can an AI agent take food orders in free text?

Yes. It maps the customer's words to the menu, applies customisation, confirms the total and time, and sends the order to the kitchen.

Does the AI agent handle allergy questions?

It answers from allergen tags; severe allergies are confirmed by a person because shared kitchens cannot always rule out cross-contamination.

Can the agent track a delivery?

Yes, from the order status, including the rider's name and estimated time.

Can the AI agent take bulk party orders?

It collects the details and suggests platters, then hands to the manager to confirm price and timing.

Does the agent speak the customer's language?

Yes, across 36 languages, and it understands dish names as customers say them.

What does the AI agent do when the kitchen is overloaded?

If the kitchen pauses orders, the agent stops taking them and tells customers when ordering resumes.

Can the AI agent send GST invoices to offices?

Yes, in the company's name with each order or monthly for credit accounts.

Can the AI agent reorder my last meal?

Yes — say "same as last time" and it confirms the items, address and total in one message.

How do I get customers ordering direct instead of through Zomato?

Every delivery customer becomes a WhatsApp contact you own. Weekly offers, live order updates and win-back messages go out on your own channel, so repeat orders stop carrying an aggregator commission.

Can I send offers to my past customers?

Yes. Build your direct customer list from WhatsApp orders and send broadcast campaigns — new dishes, weekend combos, loyalty offers — with zero platform involvement.

Written by Riya · Reputoo. Riya writes Reputoo’s guides on the WhatsApp Business API, chatbots and voice AI for Indian businesses. Prices are checked against Meta’s published rate card (25 September 2026); competitor details come from each vendor’s own site.

← Back to all articles
Put it to work

Set up WhatsApp AI agent for your business.

1,000+ businesses onboarded · Member of NVIDIA Inception · Official WhatsApp Business API, no per-message markup.

Get the details on WhatsAppNo payment needed. We will message you on WhatsApp.
+91
By sending, you agree to be contacted on WhatsApp. Prefer to write first? Message us