Case Studies | FinTech

Automating transaction validation with AI

Transaction Validation

About

A FinTech business working with Cloud Combinator on AWS. The client is anonymised at their request.

Challenge

The challenge had three focus areas, each between a raw transaction and a validated, categorised record.

Finding the reference in free text

The reference that validates a transfer is embedded in a human-written description, in no fixed position or format. The core requirement was to show that a large language model could reliably extract the matching reference number from that description across the variety of transactions the client sees.

Categorising the transfer

Beyond the reference, each transaction needs a reporting tag that classifies the transfer. The model had to return a consistent category alongside the reference, so the output is immediately useful for reporting and analysis.

Fitting cleanly into the product at scale

Whatever the model did, it had to be callable as a simple service. The solution needed to accept a transaction over an API and return structured results, in a shape that slots into the client's existing product and holds up at high monthly volumes.

Solution

A transaction is sent as JSON to an API, carrying the transaction amount and description. That call triggers an AWS Lambda function, which reads the description and uses a large language model in Amazon Bedrock, with Anthropic Claude as the reference model, to work out the matching reference and the reporting tag from the text.

100k

Transactions per month the service was designed and costed for

2

Values returned per transaction: reference and reporting tag

50%

Of the project cost contributed by AWS funding

By the numbers:

  • 100k - Transactions per month the service was designed and costed for
  • 2 - Values returned per transaction: reference and reporting tag
  • 50% - Of the project cost contributed by AWS funding
Changes

The proof of concept met its scope. It demonstrated that a language model in Amazon Bedrock could take the client's real transaction descriptions and return both the extracted reference and a reporting tag through a simple API-and-Lambda service, giving the client the evidence that AWS is a suitable base to take the capability into their product.

  • Reference extraction proven — The language model pulled the validating reference number out of free-text transaction descriptions on the client's own data.
  • Consistent categorisation — Each transaction came back with a reporting tag alongside the reference, ready for downstream reporting and analysis.
  • Product-ready service — Delivered as an API-triggered Lambda returning structured JSON, the solution is shaped to drop into the client's existing product.
  • Built to scale — The design and cost model were validated against 100,000 transactions per month, showing the approach holds at production volume.
  • Handover — The proof of concept was tested and handed over with documentation, deployed into the client's own AWS environment against a signed-off scope.

With extraction and categorisation proven on real data, the natural next step is to move the capability into production inside the client's product and to continue model development in

Amazon SageMaker, supported by a Well-Architected review of the stack. The proof of concept establishes AWS and Amazon Bedrock as a suitable foundation for that build.

AWS Stack

Amazon Bedrock

For running the large language model, with Anthropic Claude, that extracts the reference and reporting tag.

AWS Lambda

For the serverless function that reads each transaction and calls the model.

Amazon API Gateway

For accepting transactions and returning structured results over a simple API.

Amazon SageMaker

For prompt engineering and initial development before adaptation into Lambda.

Amazon S3

For storing supporting data and artefacts where required.

YOU MIGHT LIKE

Related success stories

View all case studies

Case Studies | Insights

Utilising Language Recognition, Speed, and Enhanced Security to Make Social Media a Force for Good

  • Here, we take a detailed look at how the Cloud Combinator team collaborated with another cutting-edge AI service provider that provides intelligent systems to “make social media more social” for brands and users alike.
  • Arwen AI is a UK-based startup specialising in AI solutions to manage and enhance brands’ social media interactions. Founded in 2020 by Matt McGrory, Dr. David Cole, and Joel Bailey, Arwen. AI focuses on using AI to automatically detect and remove spam, toxic comments, and other unwanted content from social media platforms.
  • The team at Arwen have three core products. ‘Moderate’ is focused on identifying and removing toxic content from social media channels. ‘Engage’ helps brands identify and engage with meaningful conversations on social media, and ‘Customize’ allows brands to apply bespoke algorithms to their channels - creating an even more effective moderation and engagement.
Read more
CONTACT US

Ready to turn AI into impact?

We'll help you spot the highest-value opportunities, reduce risk around your first AI initiative, and define a clear path to results from day one.

Why talk to us:

Outcome-driven recommendations

AWS-recognised delivery expertise

Risk-aware AI adoption

Clear next step, not a sales pitch

Start with a focused 20-minute conversation about your goals — no pressure, no commitment.

We use cookies

This website uses cookies to enhance user experience and to analyze performance and traffic on our website.

See our Privacy Policy for details.