Executive summary

Annual value at stake from AI, by use case
Indonesia ride-hailing and delivery · US$ m a year, gross, before build cost · range
Incentive targeting and pricing50-150Ads and recommendations50-110Credit scoring45-70Dispatch, batching and routing40-90Fraud and fake-order control10-40Customer service automation10-30Merchant assistants5-30Driver coach and voice tools5-25Onboarding and identity checks5-15
Source: [Kira estimates] from [GoTo FY2025], [MW 2026], [Grab Q2 2026]; logic in Chapter 3

Three reasons the order is set

The money pools are already digital. Gross value is about US$10.7 bn and the apps keep about 19% as revenue [Kira estimates] [GoTo FY2025]; a point saved on incentives is worth about US$100 m.

The leaders are already deep in. Grab runs more than 1,000 models [iNews Jan 2026]; GoTo trains its own 70-billion-parameter Bahasa Indonesia model [GoTo Q3 2025].

Driver-facing decisions are now regulated. Order allocation, pricing, incentives and suspensions must be fair and explained, with a way to object [Perpres 27/2026] [Katadata Oct 2026].

Back the uses that cut cost or lift revenue inside the apps first; treat anything that changes what a driver earns as a regulated product with an appeal desk built in.
GoTo FY2025 = GoTo Group · GoTo Q3 2025 = GoTo Group Q3 2025 results via Marketing-Interactive (30 Oct 2025): AI and automation · MW 2026 = Momentum Works food delivery report · Grab Q2 2026 = Grab Holdings · iNews Jan 2026 = iNews · Perpres 27/2026 = Presidential Regulation 27/2026 on the platform transport ecosystem · Katadata Oct 2026 = Katadata · Kira estimates = KIRA analyst triangulation

Ride-hailing and delivery sit in the most exposed corner: information-heavy work that is almost fully digital already

Exposure test: how much of the work is handling information, against how much of the proposition is already delivered through software. Ratings 1-5.

AI exposure by industry in Indonesia
KIRA rating 1-5 · information intensity against digitised share of the proposition
Hidden potentialMost exposedLow exposureDigital but simpleDigitised share of the proposition (1 low - 5 high)Information intensity (1-5)Ride-hailing and deliveryRetail bankingTelecomModern retailCold chain logisticsRestaurantsConstruction
Source: [Kira estimates]; comparison industries from KIRA reports on the same markets

Why the score is high

Every order is software. Booking, matching, price, route, payment and payout happen in the app; Grab says its models learn from more than 20 billion rides and orders [GrabX Apr 2026].

Labour sits outside the payroll. The 2-4 million drivers are partners, not staff [Grab ID Feb 2026]; AI changes their earnings more than the platform's wage bill.

The limit is the street. Traffic, rain and pickup points stay physical; models improve the guess, not the road.

Exposure is near the maximum, so the question for a newcomer is not whether AI matters here but which slice of an already-automated chain it can improve.
GrabX Apr 2026 = Grab press centre · Grab ID Feb 2026 = Grab Indonesia via Bisnis and Gizmologi · Kira estimates = KIRA analyst triangulation