Grab · AI product · Mid · 45 to 60 minutes

How should Grab summarize merchant reviews?

Grab is a Southeast Asian superapp that offers services such as food delivery, ride hailing, payments, and financial services. Answer this product management case and get it graded line by line against a weighted rubric.

Product sense
AI judgment
Trust and safety
Metrics design
Prioritization

The situation

You are a PM working on merchant discovery for Grab's food delivery experience. Customer research and internal teams keep raising the same issue, star ratings are easy to compare, but they do not tell consumers why a merchant is loved or disliked, and they do not help merchants quickly spot recurring praise or complaints. Leadership wants a recommendation on whether Grab should introduce an AI powered review summary experience on merchant pages, and whether the same review data could also create value for merchants. The opportunity is meaningful because written reviews already exist at scale, but the risks are serious. A misleading summary could hurt consumer trust, unfairly affect a merchant's reputation, or over amplify a small set of reviews. Review volume differs widely by merchant, themes change over time, and reviews can mention different aspects such as taste, packaging, portion size, value, or delivery experience. You need to propose an MVP that is useful, trustworthy, and realistic to ship without creating a large manual moderation burden.

What to produce

Write a 1 to 2 page product strategy memo.

Your answer must

Define the user problems for consumers and merchants, and choose a clear primary focus for the first launch.
Recommend the MVP experience, including what summaries appear, where they appear, and when the product should suppress or withhold a summary.
Describe the AI product approach at a high level, including how source reviews are selected, how output quality is checked, and how summaries stay current over time.
Specify success metrics and guardrails for usefulness, trust, and marketplace impact, plus a rollout or experiment plan.
Identify the top risks and mitigations, including at least one issue related to representativeness or bias and one issue related to misuse or abuse.

Constraints

You cannot assume manual review of every summary before it is shown to users.
You must use existing written reviews as the core source data for the first launch, not a new review collection flow.
Any consumer facing output must avoid claims that are not supported by source reviews, because it can affect merchant reputation.
The first launch should fit within one product area and build on existing review surfaces rather than requiring a broad platform rebuild.

Format

Use clear headings. A simple table for metrics and risks is fine.

How Grab actually played it

Grab turned written merchant reviews into short summaries that consumers can scan by topic and merchants can use as feedback. The post focuses on how the summaries are generated and kept accurate over time using feedback. It is a clean example of an AI feature that serves two sides of a marketplace.

An AI summary can serve both buyers and sellers with the same data.
Plan a feedback loop from day one to keep the output trustworthy.
Aggregate ratings hide the reasons, summaries can surface them.

Built from Automated merchant review summary system with integrated feedback by Grab Engineering.