Case Studies | SaaS B2B

Reading the room in every inbox

AI Email Sentiment Analysis PoC

About

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

Challenge

The challenge had three focus areas, each one about turning a language model from an interesting demo into a dependable step inside a live workflow.

Staying inside model service limits

A naive integration that fires every email straight at Amazon Bedrock quickly runs into throughput quotas, and once a quota is exceeded the service simply breaks. The pipeline had to pace itself so that bursts of email never pushed the model past its per-minute ceiling.

Making the output trustworthy and explainable

A bare positive, neutral or negative label is not enough to act on. The model needed carefully engineered prompts so that every classification came back in a consistent, machine-readable format and carried a short reason for the decision.

Guarding against misuse

Before any sentiment call was made, each email had to be checked for attempts at AI misuse, so that only legitimate content reached the sentiment step and anything suspect was stopped and reported back.

Solution

The flow begins in the client's own Microsoft Power Automate environment, which sends a POST request to an Amazon API Gateway endpoint carrying the header and content of the email to be analysed. That request triggers a first AWS Lambda, which places the message onto an Amazon SQS queue deliberately throttled to stay within the throughput limit of Anthropic Claude 3.5 Sonnet in Amazon Bedrock.

3-way

Explainable sentiment classification: positive, neutral or negative with a reason

100/min

Model throughput ceiling respected via SQS throttling, so the service does not break under load

~$18/mo

Projected running cost at 60 emails per day (AWS estimate, USD)

By the numbers:

  • 3-way - Explainable sentiment classification: positive, neutral or negative with a reason
  • 100/min - Model throughput ceiling respected via SQS throttling, so the service does not break under load
  • ~$18/mo - Projected running cost at 60 emails per day (AWS estimate, USD)
Changes

The proof of concept delivered against its success criteria: a working pipeline that accepts an email over an API, screens it for misuse, and returns an explainable three-way sentiment classification, all inside the client's own AWS account and ready to be called from Power Automate.

  • A quota-safe pipeline — Amazon SQS paces requests so email bursts never push Amazon Bedrock past its per-minute limit, keeping the service reliable under load.
  • Explainable results — Prompt engineering returns each classification in a consistent format with a short reason, so downstream steps can act on the output with confidence.
  • Misuse screening built in — Every email is checked for AI misuse before sentiment analysis runs, and suspect messages are stopped and reported rather than processed.
  • Deployed as code — The full stack is defined in AWS CloudFormation and stood up in eu-west-2, giving the client a repeatable, auditable deployment in their own account.
  • Handover — Cloud Combinator provided a demonstration of the service, the supporting materials and documentation, and a project handover session for the client's team.

With the proof of concept proven, the natural next steps are a Well-Architected Framework Review to harden the solution to production specification and a widening of the pipeline to

More of the client's email workflows, so the same explainable sentiment signal can support decisions across the business.

AWS Stack

Amazon Bedrock

For large language model inference, running both the misuse check and the sentiment classification.

Amazon API Gateway

For the secure POST endpoint that receives emails from Power Automate.

AWS Lambda

For the serverless functions that queue, parse and process each message.

Amazon SQS

For throttled queuing that keeps requests inside the model's throughput limit.

AWS CloudFormation

For deploying the whole stack as repeatable infrastructure as code.

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.