Generative AI Summarisation for a Global Technology Publisher
Overview
A global technology publication wanted to explore how Generative AI could improve reader engagement without disrupting its existing editorial workflow.
The publisher’s editorial platform was powered by WordPress, and the team wanted a practical way to generate concise article summaries for long-form content. The solution needed to support editorial control, integrate with the existing CMS, and allow the organisation to experiment safely with different large language models.
Vesper designed and delivered a custom AI summarisation workflow that connected WordPress with Amazon Bedrock through a secure serverless architecture.
The Challenge
The client wanted to introduce AI into the publishing process carefully and responsibly.
Key requirements included:
Generating concise summaries for long-form articles
Keeping editors in control of when summaries were created and published
Avoiding disruption to the existing WordPress editorial workflow
Allowing experimentation with different foundation models
Building a secure bridge between WordPress and AWS AI services
Creating a scalable solution without always-on infrastructure
The project needed to prove that AI could support editorial teams without replacing editorial judgement or creating unnecessary complexity.
The Solution
Vesper built a custom WordPress plugin that allowed editors to generate summaries directly from the CMS.
Editors could select an article, trigger the summary workflow, review the result, and decide how it should be displayed. This kept the editorial team in control while making AI summarisation available inside the tools they already used.
The plugin connected to Amazon Bedrock through Amazon API Gateway and AWS Lambda. This allowed the publisher to test different foundation models without rebuilding the system or locking the workflow to a single AI provider.
The architecture was designed to be secure, flexible, and scalable, with AWS handling model access, request processing, and monitoring.
Technical Delivery
The solution included:
A custom WordPress plugin for editorial use
Amazon API Gateway for secure API access
AWS Lambda for serverless request handling
Amazon Bedrock for access to foundation models
Amazon CloudWatch for monitoring and logging
IAM roles and policies for least-privilege access
The workflow was designed to support model flexibility, allowing the editorial team to test tone, length, and style across different AI models.
The Outcome
The project gave the publisher a practical and controlled way to introduce Generative AI into its editorial workflow without disrupting the CMS or reducing editorial oversight.
Editors were able to generate article summaries on demand from within WordPress, review the output, and decide how the summary should be used before publication. This meant the AI workflow supported editorial teams rather than replacing their judgement.
The solution also gave the publisher flexibility to experiment with different foundation models and compare tone, length, and quality without rebuilding the integration. Because the architecture was serverless and connected through AWS, the system could scale with publishing demand while avoiding unnecessary always-on infrastructure.
Beyond summarisation, the project created a strong foundation for future AI use cases, including translation, Q&A, content personalisation, and other editorial tools.
Why It Matters
Many publishers want to explore AI but cannot afford to compromise editorial control, trust, or workflow stability.
This project shows Vesper’s ability to implement AI in a practical, production-ready way that supports teams rather than disrupting them.
