Most advice about AI for restaurants reads like it was written by someone selling restaurant software. This guide is written the other way round: from the restaurant’s side of the bill.
The question is not “which AI tool is coolest”. It is: where does your margin actually go, and which tool puts some of it back? Start there, and most of the buying decisions get simpler.
The one number that decides everything: your true take rate
Before you buy any tool, work out what you keep from a ₹500 order on each channel you sell through. Aggregators do not publish one flat rate, and the headline commission is only the first deduction. Industry estimates and vendor calculators put the base commission at roughly 18–26% of order value, varying by city, category and contract. Then come the payment gateway fee, 18% GST charged on the commission itself, listing ads, and any discount you agreed to fund. The all-in figure commonly lands at 25–35% — and the National Restaurant Association of India has put the total burden, including advertising and promotions, as high as 40–50% for some restaurants.
Run the plain version on a ₹500 order at a 30% all-in take: ₹150 goes to the platform and its add-ons before you pay for food, staff, rent or packaging. If your food cost is ₹150 and packaging is ₹25, you have ₹175 left for everything else. On your own direct channel, the same order costs you roughly a 2% payment gateway fee — about ₹10 — plus whatever you pay for delivery. That gap, not any chatbot feature, is why direct ordering is the first job on this list.
One honesty note: those percentages are reported ranges from industry bodies and vendor analyses, not a rate card published by Zomato or Swiggy. Your contract and your monthly payout statement are the only authoritative numbers for your outlet. If a tool vendor quotes you a commission figure without that caveat, treat the rest of its pitch with the same scepticism.
The six jobs AI actually does for a restaurant
Forget product categories for a minute. A restaurant has six jobs that software can genuinely help with:
- Take orders without paying commission — your own ordering page, QR menus, WhatsApp ordering.
- Run the counter — billing, kitchen tickets, GST invoices, inventory.
- Get found and trusted — Google listing, reviews, replies.
- Sell the menu — photos, descriptions, pricing.
- Bring customers back — WhatsApp messages, loyalty, offers.
- Decide what to push, reprice or drop — which dishes make money and which just look busy.
Buy tools job by job. A tool that does one job brilliantly beats a suite that does six badly, and you can always add the next job later.
Job 1: Take orders without paying commission
This is where the money is, and it is also where the most change happened in October 2026.
Menuthere + Zao. The Kochi startup Menuthere, started in 2024, runs a zero-commission direct ordering system used by more than 1,000 restaurants in over 70 countries, most of them in India. The free plan covers 50 orders a month on a Menuthere link; Lite costs $17 a month and adds your own domain plus UPI and card payments through Razorpay or Easebuzz in India; a branded app under your own name is $99 a month. Commission is 0% on every plan — the payment gateway still charges its own fee, which is the honest distinction most “zero commission” marketing skips.
On 29 October 2026, Menuthere launched Zao, an AI agent that sets up and runs this ordering by chatting. You type the restaurant’s name, it reads your menu from a photo or PDF, and afterwards you run the shop in plain language: “add a weekend offer on biryani”, “change the delivery charge to ₹40”, “make QR codes for 12 tables”. There is an Undo button, which matters more than it sounds — an AI that edits your live menu needs a way back. Signup works in 10 languages. Zao is days old as this guide is written, so treat it as a promising launch, not a proven workhorse — but the underlying ordering system it drives is already running in real restaurants.
ONDC seller apps. The Open Network for Digital Commerce is not an app you download; it is a government-backed open protocol. You list through a seller app — magicpin, Mystore and others — and become discoverable across buyer apps. ONDC itself charges no transaction commission, which is where the famous “3%” figure comes from. The full picture: your seller app sets its own fee (published bands run roughly 1–5%), and delivery is a separate arrangement. magicpin says it has onboarded over 22,000 restaurants on ONDC. ONDC is a genuine low-cost channel, but it is not a discovery engine like Zomato — customers have to find you through a buyer app, so it works best alongside your own ordering page, not instead of it.
The new challengers. Two more commission stories are live in late 2026. Rapido’s Ownly charges restaurants no commission and a flat delivery fee to the customer, and has been reported at tens of thousands of daily orders. Flipkart’s food delivery service has proposed a base commission of about 10% — roughly half the incumbents’ typical rate — with a launch expected around the end of October 2026 and five or six cities by the end of the year. That one is still a proposal, not a channel you can list on today, so do not plan your month around it.
What about just leaving the aggregators? Do not. Aggregators still own most food discovery in India. The strategy that works is a split: keep Zomato and Swiggy for new customers, and move repeat customers to your direct channel. Every repeat order that shifts from a 30% take to a 2% gateway fee is worth roughly ₹140 on a ₹500 order — that is the entire business case for a direct ordering page, and it pays for a $17-a-month plan in a single order.
Job 2: Run the counter — POS, GST and the quote trap
If you already have a POS, do not rip it out for an AI feature. If you are buying one, the India-specific questions matter more than the AI label: GST billing that your accountant accepts, direct Swiggy and Zomato order integrations so aggregator orders land in the same queue, kitchen ticket printing that survives a Wi-Fi outage, and regional-language support if your staff need it.
Petpooja, the Ahmedabad-based market leader with tens of thousands of outlets, is the default shortlist pick for exactly those integrations. Its pricing, however, is a trap for comparison articles: Petpooja does not publish list prices, and directories and buyer guides cite wildly different bands — anywhere from about ₹700 a month to several thousand per outlet per month, plus setup fees and add-on modules for kitchen displays, QR ordering and loyalty. The honest statement is that you need a written quote for your outlet count and modules, on a typically 12-month contract, and you should compare that quote against the whole stack, not a headline number.
At the other end, WhatsApp-first systems such as Online eMenu price the whole POS-plus-ordering job at a flat ₹199 a month with monthly billing and no lock-in, according to the vendor. The lesson is not that one is better; it is that the market spans a 10x price range for the same job title, and the difference is usually integrations, contract length and support — ask about all three before signing.
Job 3: Get found and trusted — reviews are a job, not a chore
Most customers meet your restaurant on Google before they meet your food. A complete Google Business Profile — correct hours, a real menu, current photos, an order button pointing at your direct page — is free and does more than most paid tools. The AI part that earns its keep is review replies: a steady stream of thoughtful replies lifts both ranking and trust, and it is exactly the kind of repetitive writing that AI drafts well and humans should approve.
We covered this job in depth in our Google review reply guide for Indian SMBs, including the DIY-versus-subscribe math and the compliance lines that matter: never gate reviews behind a good experience, never pay for them, and remember that India’s BIS standard IS 19000:2022 sets expectations for genuine, verifiable reviews. Two rules of thumb from that guide apply here: AI drafts, a human approves every reply, and a reply to a bad review is worth more than ten replies to good ones.
Job 4: Sell the menu — photos that help, photos that lie
AI food photography is the most oversold tool in this whole guide, so here is the line: AI can clean up a bad photo of a real dish. It must not invent a dish you do not serve. Customers who receive food that looks nothing like the photo do not complain about the AI — they leave a one-star review, which sends you back to Job 3 with a problem you created.
The legitimate uses are real: background cleanup for menu photos, consistent lighting across a 60-dish menu, and quick Hindi or English descriptions from a dish’s ingredients. Our AI image generators comparison covers the tools and their India pricing if you want to do this yourself. The practical test: if you would be embarrassed serving the photographed dish next to the photo, do not publish the photo.
Pricing deserves one line too. Many restaurants price aggregator menus 17–22% higher than dine-in menus to absorb commission — a common, openly discussed practice. If you do this, keep your direct-channel menu at the real price. The price gap itself becomes a reason for customers to order direct.
Job 5: Bring customers back — WhatsApp is your cheapest channel
Your best customers are the ones whose numbers you already have. WhatsApp ordering and messaging is where small Indian restaurants get the highest return from AI, because the customer is already on the app and the message actually gets read.
The no-code path — a chatbot that takes orders, answers “is paneer tikka available today?” and sends order updates — is covered step by step in our WhatsApp chatbot guide for Indian businesses, including the per-message costs that decide whether it pays. The one number to remember from the pricing side: WhatsApp Business API messages in India cost roughly ₹1.09 for a marketing message and ₹0.145 for a utility or service message, with 1,000 free service messages per number per month — full detail in our WhatsApp Business API pricing guide. At those rates, a “your usual biryani is 20% off today” message to 500 past customers costs less than one aggregator order’s commission.
One caution from the same guide: WhatsApp messages need opt-in, and template rules apply. A customer who ordered once has not necessarily agreed to weekly promotions. Ask once, respect a “no”, and your channel stays deliverable.
Job 6: Decide what to push, reprice or drop
This is the newest AI job and the least mature. The idea is simple: instead of a dashboard of numbers, you get a short daily list — push the high-margin dish that is selling slowly, reprice or drop the dish that loses money, watch the aggregator that cost you most this week, message the guests worth winning back. Online eMenu builds exactly this into its AI restaurant system, and Menuthere’s Zao answers “how do I increase my sales?” from the restaurant’s own numbers.
Treat these recommendations as a second opinion, not an order. An AI that sees only sales data will tell you to drop the low-margin dish that brings families in on Sundays. You know the regulars; it knows the spreadsheet. Use both. The decision framework that works: let the AI surface the candidates, let the owner decide, and review the decision a month later against actual numbers.
The starter stack: what to buy first, by restaurant type
| Restaurant type | First buy | Second buy | Skip for now |
|---|---|---|---|
| Single outlet, dine-in + takeaway | Free direct ordering page (Menuthere free plan or equivalent) + Google Business Profile cleanup | WhatsApp ordering + review replies | Branded app, full POS replacement |
| Cloud kitchen, delivery-first | Direct ordering + ONDC seller listing | WhatsApp repeat-customer messages | Dine-in features (QR table ordering, KDS hardware) |
| 2–5 outlet small chain | POS with aggregator integrations + written quote comparison | Central menu management + review replies across outlets | Custom AI development — buy, do not build |
The total monthly cost of the first column is ₹0 plus a payment gateway fee. That is deliberate. Most restaurants do not have a tool problem; they have a channel problem — too many orders flowing through the most expensive channel. Fix the channel first, then buy tools for the jobs that remain painful.
The traps
GloriaFood’s shutdown. If your ordering runs on GloriaFood, migrate now. The Oracle-owned service stopped new sign-ups in April 2026 and reaches end of life on 30 April 2027. Export your menu, your customer list and your domain mapping before you need them in a hurry.
“Lifetime deal” ordering software. A one-time ₹5,000–₹10,000 “lifetime” tool with no company behind it is a subscription that ends when the seller disappears — and your menu goes with it. Prefer tools with a monthly plan you can cancel; the cancellation option is the feature.
AI photos of food you do not serve. Covered above, worth repeating: one viral “this is not what I ordered” post costs more than a year of honest photography.
Fake reviews, bought or AI-generated. Beyond the ethics, they are detectable and they poison the trust that Job 3 is building. India’s review standard exists for a reason; one exposed batch can define your listing.
Paying for AI before fixing the basics. An AI menu-pricing tool on top of a Google listing with wrong hours and a dead phone number is decoration. Hours, phone, menu, photos, order button — in that order.
The bottom line
The AI tools worth buying for an Indian restaurant in 2026 are mostly not exotic. A direct ordering page that skips a 25–35% take. A POS that talks to the aggregators and prints a GST invoice. Review replies that actually get written. WhatsApp messages to people who already like your food. A short daily list of what to push and what to drop.
Start with the channel, not the chatbot. Move one in five orders from an aggregator to your own page, and you have funded every other tool on this list — with money that used to leave the building on every single order.
Sources
- Menuthere pricing and Zao launch: menuthere.com (Free $0/50 orders a month; Lite $17 a month; Branded App $99 a month; 0% commission; Razorpay/Easebuzz for India); Tribune India / ANI, 29 October 2026 (Zao launch, 1,000+ restaurants, 70+ countries, 10 languages, GloriaFood end of life 30 April 2027).
- Aggregator commission ranges: Online eMenu commission calculator and 2026 guides (base 20–24%, all-in 25–35%); Moneycontrol/BigGo Finance reporting of NRAI figures (base typically 20–25%, total burden with ads and discounts 40–50%); Flipkart ~10% proposed commission and Rapido Ownly zero-commission reporting (StartupFeed, BigGo Finance, October 2026).
- ONDC: magicpin ONDC guide (22,000+ restaurants; seller-app fees 3–5% framing); BuildByRaviRai ONDC seller maths, October 2026 (ONDC charges no transaction commission; seller-app fees roughly 1–5%; delivery separate).
- Petpooja: Online eMenu POS comparison, July 2026, and buyer directories (no published list pricing; cited bands conflict widely; 12-month contracts typical; aggregator API and GST integrations).
- KnowledgeSense’s own verified guides: Google review reply automation (October 2026), WhatsApp chatbot builder (October 2026), WhatsApp Business API pricing (September 2026), AI image generators (September 2026), paying for AI tools in India (October 2026).
This article was researched and written with AI assistance, and every price, date and claim above was checked against the sources listed. Prices and commission rates change — verify the current figure on the vendor’s page or your own payout statement before signing anything.
