What challenges did Trustly face before AI implementation?

Authorize or decline payments in real time at scale while improving fraud prevention. Prior systems and outsourced fraud workflows were not optimized for large data volumes and evolving fraud threats, leading to slower risk assessments and higher fraud attempts and false positives.

What was the impact of the AI solution?

Reduced fraud attempts by 52% while maintaining 99.99% system availability. Reduced false positives by 43%. Improved transaction risk assessments by 75%. Reduced operational costs by 40%. Cut model deployment time from weeks to hours.

Rank#594of 3,449

Source

Scaled productionEvidence: Medium65/100

Accelerating Payment Approvals by 75% using Amazon Bedrock — Trustly

Use case typeFraud detectionUpdated Jul 15, 2026

Trustly is a global payment processing company that manages transactions between more than 110 million customers and merchants across more than 30 countries. The company migrated from private data centers and outsourced tools to AWS to improve fraud detection, transaction approval speed, visibility, and scalability.

Organization: Trustly
Industry: Finance
Location: Sweden
Published: July 2026

Reported outcomes

−52%

fraud attemptsRisk, reliability & safety

−43%false positives−75%transaction risk assessments−40%operational costs

Strategic outcomes

Speed & agilityCut model deployment time from weeks to hoursScale & capacityMaintained high availability at payment-processing scaleCustomer experience & trustImproved merchant visibility and reporting

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Fraud attempts: 52% decrease

AWS Solutions Case StudyJul 15, 2026Customer storyExplicit claimMedium evidence strength

reduced fraud attempts by 52 percent

Normalized claim

False positives: 43% decrease

AWS Solutions Case StudyJul 15, 2026Customer storyExplicit claimMedium evidence strength

reduced false positives by 43 percent

Normalized claim

Transaction risk assessments: 75% decrease

AWS Solutions Case StudyJul 15, 2026Customer storyExplicit claimMedium evidence strength

transaction risk assessments are 75 percent faster

Normalized claim

Operational costs: 40% decrease

AWS Solutions Case StudyJul 15, 2026Customer storyExplicit claimMedium evidence strength

reducing operational costs by 40 percent

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Trustly
Provider: AWS
Maturity: Scaled Production
Linked source: AWS Solutions Case Study

Authorize or decline payments in real time at scale while improving fraud prevention

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Fraud detection
2Risk assessment
3Operational analytics

Authorize or decline payments in real time at scale while improving fraud prevention.
Prior systems and outsourced fraud workflows were not optimized for large data volumes and evolving fraud threats, leading to slower risk assessments and higher fraud attempts and false positives.

Trustly migrated data and workloads to AWS, using Amazon RDS for relational data management.
It built and ran fraud detection ML models in Amazon SageMaker.
It used Amazon Bedrock to categorize banking activity and improve model accuracy for fraud detection.
It adopted Amazon Q to provide dashboards and AI-assisted workflows for risk and fraud prevention teams.

Technologies

Amazon Bedrock Amazon SageMaker Amazon RDS Amazon Quick

Reduced fraud attempts by 52% while maintaining 99.99% system availability.
Reduced false positives by 43%.
Improved transaction risk assessments by 75%.
Reduced operational costs by 40%.
Cut model deployment time from weeks to hours.

Architecture

Trustly migrated from private data centers and outsourced tools to AWS, storing data in Amazon RDS, building fraud detection models in Amazon SageMaker, using Amazon Bedrock for transaction categorization, and using Amazon Q for analytics dashboards and AI-assisted actions for fraud teams.

Sources & evidence1

Evidence: Medium65/100Evidence strength