Grab
AI assistant for merchant advice and activation
In 2025, Grab's AI merchant assistant surpassed one million interactions in under five months in Indonesia, with 96.1% satisfaction and advertising efficiency up 11% among the merchants who use it.
Objective
Give every merchant, even small and without a marketing team, an always-on advisor that answers their operational questions, creates their advertising campaigns, and updates their menu, to lift their revenue and their ad spend on the platform.
The deployment
In April 2025, Grab launched an AI assistant for its merchants, presented as a business advisor available around the clock in the GrabMerchant app. The merchant asks it operational questions (how to sell more, adjust the menu, run a promo) and the assistant responds with personalized recommendations, can create advertising campaigns, and update the menu. It can also take the initiative and suggest actions. In Indonesia, where the tool was deployed at scale, merchants exchanged with the assistant more than one million times in under five months, with a high satisfaction rate. Grab measures an improvement in advertising efficiency among the merchants who use it.
Results Proof C
Official Grab press release (T1) on the launch and architecture, and established regional press (Marketing-Interactive, T3) reporting the usage and efficiency figures in Indonesia. Consistent sources. The 11% advertising efficiency is a figure reported by Grab, not an audited result, which keeps it at C.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
The stack in detail
- llm OpenAI Grab's model partner (collaboration announced in May 2024, expanded in 2025); exact model not named publicly
- llm Anthropic Grab's strategic partner on agentic AI solutions
- plateforme App GrabMerchant (integration in-house) Merchant back-office into which the assistant is integrated, with access to each merchant's sales, campaigns, and menu
- outil GrabAds Grab's advertising platform on which the assistant directly creates the merchants' campaigns
How it runs, concretely
For ops teams-
1Merchant question merchant
The merchant opens the chat in GrabMerchant and asks a business question or an action request.
-
2Reading the context AI
The assistant reads the merchant's sales, campaign, and menu data to answer their specific case.
-
3Recommendation or action AI
It suggests an action (promo, menu adjustment) or executes it, such as creating an advertising campaign.
-
4Proactive follow-up AI
The assistant comes back to the merchant with suggestions to improve their sales.
The merchant's operational data (sales, campaigns, menu). Without an order and campaign history specific to each merchant, the assistant cannot personalize its advice.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- Order and sales history per merchant
- Advertising campaign data
- Menu or catalog structure per merchant
Org prerequisites
- AI transparency rule toward the merchant
- Connected action tools (campaign creation, menu editing)
- Measurement loop on ad efficiency and adoption
Possible stack
- Generative LLM (OpenAI, Anthropic, or an open model)
- Tools layer connected to the merchant back-office
- Integration with the advertising platform
The plan, step by step
- Step 1List the 10 most frequent merchant questions and actions (support, account managers) and check that sales, campaigns, and catalog are accessible per merchant.Deliverable: Prioritized functional scope and validated data access
- Step 2Connect the LLM to the merchant's operational data in read mode, then to the first two actions in write mode: campaign creation and menu editing.Deliverable: Assistant able to answer the merchant's case and execute two actions
- Step 3Launch a pilot on a market or a segment of merchants, with AI transparency and guardrails on the actions (confirmation before execution).Deliverable: Pilot in production with conversation logging
- Step 4Measure adoption, satisfaction, and advertising efficiency of the merchants using it against a control group.Deliverable: Quantified adoption / satisfaction / ad efficiency review
- Step 5Add proactive follow-ups (suggestions sent by the assistant) and extend to the next markets.Deliverable: Proactive assistant deployed beyond the pilot
First step: List the 10 most frequent merchant questions and actions, and first connect campaign creation and menu editing.
The same pattern, proven in other industries
Sources
- S1 Grab deploys agentic AI to empower merchants and driver partners Primary archive pending
- S2 Grab's AI assistant surpasses 1m merchant interactions in under five months Established press archive pending
- S3 Grab Reports Fourth Quarter and Full Year 2024 Results Primary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.