Case Studies | FinTech

Completing the mortgage Fact Find from a meeting recording

PoC - AI Fact Find Answering

About

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

Challenge

The challenge had four focus areas, each a step between a recorded conversation and a completed, trustworthy Fact Find.

Speech into usable data

The starting point is an audio or video recording of a meeting. That has to become clean, structured text before any language model can reason over it, so

Answering many questions accurately

A Fact Find is not one question but many, spread across fifteen sections. The core requirement was to show that a large language model within Amazon Bedrock could answer that volume of questions from the transcript with high accuracy, rather than only handling a simple prompt.

Turning the meeting into actions

Beyond the answers, advisers need the next steps: obtaining policy documentation, researching remortgage options, arranging protection quotes, scheduling follow-up meetings. The solution had to read the whole conversation and produce a clear summary of the action items to send back to the user.

Handling sensitive data safely

Client financial data is highly sensitive, so the pipeline had to run inside a private, controlled AWS environment and include a guardrail step to ensure no content designed to misuse the language model was passed through for processing.

Solution

A recording is uploaded to Amazon S3 using a presigned URL. That upload triggers a Lambda function which sends the file to Amazon Transcribe, with a SQS queue placed in front to keep within Transcribe usage limits. The finished transcript is written back into an uniquely identifiable folder in S3 that holds everything needed to process that meeting.

2 of 15

Fact Find sections processed end to end to prove the approach

45 min

Advisory recordings turned into structured Fact Find answers

By the numbers:

  • 2 of 15 - Fact Find sections processed end to end to prove the approach
  • 45 min - Advisory recordings turned into structured Fact Find answers
Changes

The proof of concept met its scope. By deliberately processing two of the fifteen Fact Find sections end to end, we demonstrated that Amazon Transcribe and Amazon Bedrock together can take a recorded meeting and return accurate, structured answers plus a summary of follow-up actions, giving the client the evidence to decide on a move into full production.

  • Transcription proven — Amazon Transcribe converted the client's meeting recordings into text of a quality suitable for a language model to work from.
  • Sectioned answering — A per-section Lambda and prompt design let Amazon Bedrock answer a defined slice of the Fact Find, an approach that scales cleanly to the full fifteen sections.
  • Action summaries delivered — The pipeline combined the section results and produced a summary of the meeting's action items, returned to the user at the end of the run.
  • Secure by design — The workflow ran inside a VPC with a Bedrock-based guardrail step, keeping sensitive client data contained and screening input before processing.
  • Handover — The proof of concept was delivered with the architecture and documentation needed for the client to judge production readiness against the original scope.

With two sections proven, the clear next step is to extend the same pattern across all fifteen sections of the Fact Find and host the pipeline in the client's own AWS environment for production use. The proof of concept establishes that AWS and Amazon Bedrock are a suitable foundation for that build.

AWS Stack

Amazon Transcribe

For converting recorded advisory meetings into text a language model can process.

Amazon Bedrock

For answering the Fact Find questions, screening input and summarising action items.

Amazon S3

For holding each meeting's recording, prompts, section answers and combined output.

AWS Lambda

For the serverless steps that move data through transcription, answering and summarising.

Amazon SQS

And Amazon SNS for controlling throughput and coordinating the fan-out and completion steps.

Amazon VPC

For keeping the sensitive data pipeline within a private, controlled network boundary.

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.