UruguayFinance and insurance2025 – ongoingdLocal
AI agent for merchant website compliance reviews
dLocal finished up to 75% of merchant website reviews with no human intervention in controlled evaluations
Problem
dLocal, a cross-border payments company, had to review thousands of merchant websites each month for prohibited goods and services. Its policy covers more than 15 categories and more than 100 subcategories, and merchants change their offerings over time. A generative AI tool from 2023 still left many cases to manual review.
Approach
dLocal built the workflow with the AWS team on Amazon Quick Automate. A UI agent navigates merchant product pages in several languages and checks items against the prohibited categories. Each website is a case, processed in parallel batches, and is classified as approved, rejected or website inaccessible, with reasoning. Rejected cases go to human review with evidence and screenshots.
Results
Takeaway
dLocal kept people on rejected and high-risk cases and let the agent clear the routine ones. The company aims to extend automated coverage to most compliance cases.
Categories
- Industry
- Finance and insurance
- Business function
- Finance, risk and legal
- Innovation type
- Process innovation
- AI technology
- AI agentsGenerative AI
- Adoption
- Partner co-development
- Evidence
- Vendor case study
Sources
Source summaries
The post, co-written by AWS staff and a dLocal architect, describes a UI agent in Amazon Quick Automate that reviews merchant websites for prohibited products. It reports that up to 75% of reviews finished with no human intervention in controlled evaluations. Rejected cases go to human review.
Full details are in the original sources.
