How KOHO transformed from a financial app to an AI leader using AWS | Case Study

Serving over 2 million customers across Canada, KOHO offers transparent financial products. The company’s infrastructure is built entirely on AWS, and KOHO consistently searches for new technologies to help enhance customer experiences and improve operational efficiency.

KOHO’s leadership recognized an opportunity to move beyond using AI as isolated tools. Given its experience with other foundational AWS services, the company knew that using AI to build out capabilities, rather than specific solutions, could vastly improve team productivity while making sure that customer data security was a primary consideration. “Our goal was to provide KOHO employees with better, easier, and more secure access to large language models,” says Manikandan Paramasivan, senior staff architect for data and AI at KOHO. The transformation to holistically embed AI capabilities throughout every aspect of its FSI operations needed to happen, however, without compromising the strict financial services data privacy and security standards. At the same time, the solution needed to be flexible enough to adapt as AI technology continued to evolve.

To achieve this AI integration, KOHO turned to AWS to provide both the scalability to run AI workloads efficiently and the tool set needed to build generative AI applications, all while maintaining a strict data governance framework. “Using AWS services for our AI workloads made complete sense because it provided the flexibility that we’d need for AI integration, and we were already using AWS services for security,” says David Kormushoff, vice president of technology at KOHO.

Working alongside the AWS team, KOHO developed Kortex AI, a generative AI solution designed to provide secure, scalable access to AI across the entire organization. At the core of the solution is Amazon Bedrock, a comprehensive, secure, and flexible service for building generative AI applications and agents. By building on Amazon Bedrock instead of using a third-party solution, KOHO can be confident that all AI activity stays within its AWS infrastructure: Sensitive customer data is processed entirely in a controlled environment while foundational models are accessed through secure APIs. To meet financial service requirements, Kortex AI enforces additional safeguards, such as centralized authentication and logging, model approval lists, and restricted access to data sources. These controls are particularly important for the financial industry, where sensitive customer transactions and personal information must remain protected to adhere to strict regulatory requirements.

To support Kortex AI’s data requirements, KOHO uses Amazon Aurora, which provides high performance and availability at global scale, as its vector database. This service facilitates sophisticated retrieval augmented generation capabilities for use cases requiring contextual data enrichment. The solution also uses Amazon Elastic Kubernetes Service (Amazon EKS), a service used to build, run, and scale production-ready Kubernetes applications. Using Amazon EKS provides KOHO with the container orchestration needed to manage AI workflows, API interactions, and user interfaces at scale.

Employees across KOHO can access large language models through a web-based interface for research, analysis, and general productivity tasks, while developers receive AI coding assistance directly in their integrated development environments. The company also uses AI to schedule and prepare for meetings, helping to boost efficiency. “Using AWS, we weaved AI into our workflows to help create meeting notes and gather information before talking to developers,” says Kormushoff. “This helps us to have better, more productive conversations.”

 

Beyond these organization-wide capabilities, Kortex AI powers purpose-built solutions for KOHO’s core financial services operations. Using Kortex AI, KOHO can validate and submit anti–money laundering suspicious-transaction reports three times faster, reducing money laundering investigation time by days and helping to prevent potential regulatory fines. Security event resolution time has dropped 50 percent, and analysts can query data using natural language, saving hours on market- and customer-trend analysis.

KOHO uses AI for a host of additional use cases, including to rapidly analyze and implement customer-requested features. For example, after Kortex AI identified “dark mode” as one of the most requested features, the enhancement was built in 2 days and production-ready within weeks. The company is also using AI for rental lease validation, decreasing the time and effort needed to validate rental leases for KOHO’s “Rent Reporting” product. Employees also use AI to assist with presentation and document creation for investor presentations and regulatory reporting.

While developer productivity remains a strong proof point of the solution, with 120–140 percent higher pull request throughput and 66 percent more frequent deployments, Kortex AI’s real impact lies in helping employees make faster, smarter decisions across the business.

KOHO’s journey illustrates the benefits of financial institutions’ holistic adoption of AI. Looking ahead, the company plans to expand its AI capabilities even further by implementing Amazon Bedrock AgentCore—a service used to deploy and operate AI agents securely at scale—to facilitate complex, multistep decision-making across business processes. “With all of these improvements, we are getting more efficient and deploying more features, but the quality checks that we have in place remain the same,” says Kormushoff. “This maintains consistency and quality in our system—there’s no degradation in service quality from deploying more AI.”