What happened
Google has brought Gemini's task automation to India. You ask Gemini to order food, book a ride or reserve a table, and it does the tapping for you inside the app. Google announced the expansion to India, the UK and Japan on 10 November, and its India blog said on 12 November that the feature is now available here.
- Gemini opens the app on a virtual screen on your phone, searches, picks items and fills the cart. It stops before the last step, so you confirm and pay.
- It only works on recent premium phones: the Pixel 10 and 11 ranges (not the Pixel 10a) and Samsung's Galaxy S26 and Z8 ranges, with a Motorola phone coming. The phones need at least 12GB of RAM and Android 17, according to Google's Android site as reported by PPC Land.
- It's a beta for users aged 18 and over. In India it works in English. No subscription is needed to start, though there are daily limits.
- Google hasn't said which apps it works with. Android Authority found a list on Google's servers with 114 apps worldwide and 9 in India: Zomato, Blinkit, Rapido, BookMyShow, ixigo, McDelivery, KFC India, Burger King India and Starbucks India.
- In its tests, ordering worked on Zomato and Blinkit. Asked to use Swiggy, Gemini replied: “Swiggy is not available for automation”. It said the same for Zepto.
- Google sends the list from its servers, so it can change at any time. It isn't known whether the missing apps chose to stay off or simply haven't been added yet.
For a product manager, this matters beyond the gadget news. When an assistant does the browsing, it may skip the screens your app makes money from: offer banners, sponsored listings and the “add a drink?” prompt before checkout.
The case
You're a product manager at a food delivery app in India. Gemini can now order on a rival app, but not on yours. Your CEO asks: should we push to get on Google's list, build our own integration with Gemini, or stay off?
1Clarify
- Is being off the list our choice or Google's? If Google simply hasn't added us, the question is whether to ask. If we opted out, someone already weighed this, and I'd want to know why.
- How many of our users have an eligible phone? Only premium phones qualify today, so the reach could be tiny.
- How much of our money per order comes from screens before the cart? Ads and upsells are what an assistant might skip, so this sets the cost of joining.
- Do we still own the order? If the order runs inside our app, we keep the customer, the data and the membership. That's very different from an order placed somewhere else.
2Structure
- Reach: how many orders could come through an assistant, now and in two or three years?
- Unit economics: what does one assistant order earn compared with a normal one?
- Control: who owns the customer relationship, the ranking of restaurants and the data?
- Experience risk: wrong orders, refunds and support calls when an assistant gets it wrong.
Then compare three options against those four areas: join Google's list, build a proper integration (Google opened a preview of a platform for this in October, PPC Land reports), or stay off.
3Analyse
How big is this today?
All the numbers below are assumptions for the case, not any company's real data.
| Step | Assumption | Result |
|---|---|---|
| Monthly ordering users | Our app | 1 crore |
| On an eligible phone | 2% of users | 2 lakh |
| Use Gemini to order in a month | 10% of those | 20,000 |
| Assistant orders a month | 4 each | 80,000 |
So assistant orders would be a very small slice of our total business today. The question is how much each choice costs per month, and how fast that grows.
The cost of joining, and the cost of staying off
Say each order earns us ₹40 from commission and fees, plus ₹12 from ads and upsells. Assume the assistant skips half the ad and upsell value, so we lose ₹6 on every assistant order.
- If we join: 80,000 orders × ₹6 = ₹4.8 lakh a month in lost ad and upsell revenue.
- If we stay off: some people who like ordering through Gemini will use the rival instead. If 25% of those 80,000 orders move, we lose 20,000 orders × ₹52 = ₹10.4 lakh a month.
Costs per month on the assumptions above. A lost order loses the full ₹52; an assistant order loses only the ₹6 of skipped ads and upsells.
The break-even point is useful. Joining costs ₹4.8 lakh. Staying off costs 80,000 × ₹52 × the share that switches, which is ₹41.6 lakh × that share. The two are equal when the share is ₹4.8 lakh ÷ ₹41.6 lakh, about 12%. If more than roughly one in eight assistant users would go to the rival rather than order from us by hand, joining is cheaper.
What the maths leaves out
Control matters as much as the money. With screen automation, the order still happens inside our app, so we keep the customer, the payment and the membership. A built integration would let us decide what the assistant sees, such as our offers and our membership price. The real risk would be an assistant that picks the app for the user. Then the restaurant listing we rank today would be ranked by someone else.
4Recommend
I recommend we ask Google to add us now, and at the same time scope our own integration, so that we decide what the assistant sees.
First, on our assumptions, staying off costs more than joining once about 12% of assistant users would switch to a rival. Switching is easy, because the user only has to name a different app. Second, the order still runs in our app, so we keep the customer, the data and our membership. Third, joining while the numbers are small lets us learn cheaply: we can compare assistant orders with normal ones on basket size, refunds and repeat rate.
The main risk is wrong orders. If Gemini adds the wrong item and the customer blames us, refunds and support costs go up. I'd track the refund rate on assistant orders weekly and set a limit for raising it with Google.