Scaling Transaction Categorisation at 365 Business Finance with AWS

365 Business Finance implemented a scalable solution using AWS to automate the categorisation of bank transactions. By combining automation with a user-friendly interface for manual adjustments, the system improved efficiency, accuracy, and set the stage for future machine learning integration.

Problem Statement

The 365 Business Finance faced critical challenges in managing an overwhelming volume of bank transactions, which required time-consuming and error-prone manual categorisation. Their current system lacked the flexibility to adapt to evolving transaction patterns, causing delays and inefficiencies in identifying transactions from specific payment processors. The need for a more intelligent, automated system became urgent as human errors increased, underwriters were slowed down by inefficiencies, and categorisation rules became outdated.

To address this, the client sought a solution that would automate much of the categorisation process while still allowing underwriters to override and correct automated decisions when necessary. By implementing an intelligent, rules-based system that learns from these manual interventions, 365 Business Finance aimed to enhance their operational efficiency, reduce errors, and improve adaptability.

The ultimate goal was to transition to a machine learning-driven solution that could continuously improve from human input, helping the client maintain high accuracy with less manual involvement. This transformation would empower the business to scale efficiently while minimising the risks of human error.

Proposed Solution

For 365 Business Finance, we collaborated to design a cloud-based solution that automates and streamlines the categorisation of bank statements. Given the large volume and complexity of transactions, the client  needed a system capable of classifying transactions efficiently, while also providing underwriters the flexibility to manually adjust classifications when necessary. Our solution leverages AWS cloud services to deliver both automated categorisation and a robust, intuitive user interface (UI) for manual intervention.

The categorisation process is managed through a custom-built UI, designed specifically for underwriters. Rather than relying solely on automated text-matching rules, the system allows underwriters to manually review and categorise transactions. Powered by Amazon Cognito for secure authentication and AWS Amplify for seamless deployment, the UI ensures a smooth, responsive experience. Each manual adjustment is captured and incorporated into the system, refining future categorisation rules and enhancing overall accuracy.

AWS Lambda functions handle core processing, applying categorisation rules and storing the results in Amazon Aurora, while Amazon Textract extracts key information from transaction descriptions to support more sophisticated categorisation. Monitoring and logging through AWS CloudWatch and CloudTrail provide real-time insights and security, ensuring the system is reliable and scalable.

This solution empowers 365 Business Finance to handle growing transaction volumes efficiently by combining automation with manual intervention where needed. It is also designed for future growth, with plans to incorporate machine learning to further improve categorisation accuracy over time, reducing the need for ongoing manual adjustments.

Outcomes of Project & Success Metrics

Enhanced Efficiency and Accuracy

The 365 Business Finance now benefits from an automated system that significantly reduces the need for manual intervention in transaction categorisation. The user-friendly interface allows underwriters to easily manage exceptions, while their corrections continuously improve the system’s accuracy. This streamlined approach has allowed the team to focus on more complex tasks, leading to improved overall efficiency.

• Significant reduction in manual categorisation efforts
• High accuracy in automated transaction categorisation
• Faster categorisation turnaround time

Scalability and Future-Readiness

The AWS-powered solution has given the client the flexibility to scale as their transaction volume increases, without sacrificing performance. With a strong foundation for future machine learning integration, the system is positioned to continuously learn and adapt to new patterns, offering even greater efficiency over time.

• Ability to handle larger transaction volumes seamlessly
• Continuous uptime during high-traffic periods
• Strong framework for integrating machine learning to enhance automation

Improved Underwriter Experience

The custom-designed UI offers underwriters a smooth platform to override and adjust categorisations as needed. This enhancement has improved productivity and satisfaction, allowing underwriters to engage in more meaningful work instead of getting bogged down by manual processes.

• Noticeable reduction in time spent per manual adjustment
• Improved underwriter satisfaction and productivity
• Continuous integration of manual corrections into system rules to improve future accuracy