All case studies

eexpand · 2024 · Product Manager

An LLM writing assistant

A button that writes a personalised message in one click, to get companies talking to each other.

Context and problem

Engagement in conversations between companies on the Trade Club was weak. We wanted to make it easier for users to start and continue an exchange.

My role

I led the project end to end: functional specification, wireframes in PowerPoint, Jira tickets, following development with the engineers, then deploying and testing in production.

What I did

  • Added a “Write a message with AI” button that generates a tailored, coherent message in one click.
  • Configured the OpenAI API assistant behind the button, with rules for writing one specific type of message.
  • Included context on the sending and receiving companies in each request, so the message is personalised.
  • Anonymised company information before it reached the model, and added it back automatically after the response.
  • Kept the generated message editable, with a one-click return to the original text.
  • Did extensive prompt engineering and testing, then presented the results to the Technical Steering Committee, in front of bank representatives.

Result

The Technical Steering Committee approved the feature, and it was then released to users.

Tools

  • OpenAI API
  • Prompt engineering and testing
  • Jira
  • Wireframes (PowerPoint)