Omise AI
AI-powered efficiency
Back
Products
For Merchants
For Platforms and Banks
Professional Services
Resources
Help center
AI adoption is accelerating across industries—and financial technology is no exception. From personalized customer service and fraud detection to predictive analytics and workflow automation, AI is changing the game for the payments sector.
As a leading payment solutions provider in Asia-Pacific, Omise is actively harnessing the power of AI to drive innovation in payments. One of our latest developments is Omise Agent: a personal AI assistant designed to make using Omise products easier than ever.
Instead of spending time digging through documentation, merchants can now ask Omise Agent directly for help with API setup and receive step-by-step guidance instantly. It can also handle repetitive tasks, helping businesses save time and stay focused on what matters most.
But while Omise Agent is designed to work simply, building it is anything but. At AWS Summit 2025, our Director of Engineering, Sylvain Dormieu, shared key lessons from the development process—highlighting our hands-on experience with Amazon Bedrock.
In the early stages, the team faced several challenges. Although the Omise knowledge base offered comprehensive data, the AI assistant’s responses were unstructured, lacked key information, and were often inconsistent—even when answering the same two simple questions.
With the help of AWS Bedrock, a fully managed AWS service that enables teams to build and scale generative AI applications using foundation models from leading providers, we gained the tools needed to address these challenges more effectively.
AWS Bedrock offers high-quality AI models like Claude Sonnet 3.7, along with features that support workflow orchestration, conversational handling, guardrails, and—most importantly—evaluation.
Sylvain highlighted that “Getting started with AWS Bedrock is straightforward, but achieving high-quality results takes thoughtful content development and evaluation.”
At Omise, we assess AI performance through both knowledge base evaluations and agent end-to-end evaluations, measuring the quality of generated responses based on metrics such as correctness, relevance, completeness, and fluency. In short, we use AI to detect its own knowledge gaps and refine its responses in a continuous feedback loop — accelerating improvement beyond what human experts can achieve.
As we continue refining Omise Agent, our focus is on building AI tools that are smart, reliable, and built for real-world use. With Amazon Bedrock, we’re pushing what’s possible in AI-powered payments. And this is just the beginning.
