Back office ops · Production

Thomson Reuters builds Open Arena, an enterprise LLM playground, in under 6 weeks with AWS

The problem

Thomson Reuters needed a safe, enterprise-grade platform to let employees without coding backgrounds experiment with LLMs and discover AI use cases for their daily work and products.

Workflow diagram · grounded in source
1
Employee accesses tile interface
trigger
“Open Arena adopts a user-friendly interface, designed with pre-set enabling tiles for each experience”
2
RAG retrieves relevant chunks
ai_action
“we have a retrieval augmented generation (RAG) pipeline in place, which will fetch the most relevant content against the query. In such pipelines, documents are split into chunks and then embeddings are created and stored in OpenSearch. …”
3
LLM generates response
ai_action
“The retrieved best match is then passed as an input to the LLM along with the query to generate the best response”
4
Answers delivered to employees
output
“Open Arena has been developed to get quick answers from several sets of corpora, such as for customer support agents, solutions to get quick answers from websites, solutions to summarize and verify points in a document”
5
New use cases surface and feed back
feedback_loop
“Open Arena's launch led to an influx of new use cases, effectively harnessing the power of LLMs combined with Thomson Reuters's vast data resources”
Reported outcome

Open Arena reached over 1,000 monthly internal users within a month of launch, with an average interaction time of 5 minutes per user, and generated an influx of new AI use cases across Thomson Reuters's global teams.

Reported metrics
Monthly internal usersover 1,000
Average interaction time per user5 minutes per user
Platform build timeunder 6 weeks
Time to reach 1,000 monthly usersunder a month
Reported stack
Amazon API GatewayAmazon S3Amazon CloudFrontAWS CodeBuildAWS CodePipelineAmazon CloudWatchOpenSearchAmazon BedrockAmazon SageMaker JumpstartHugging FaceRAGbitsandbytesFlan-T5-XLOpen AssistantMPTFalcon
Source
https://aws.amazon.com/blogs/machine-learning/how-thomson-reuters-developed-open-arena-an-enterprise-grade-large-language-model-playground-in-under-6-weeks?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Open Arena reached over 1,000 monthly internal users within a month of launch, with an average interaction time of 5 minutes per user, and generated an influx of new AI use cases across Thomson Reuters's global teams.

What tools did this team use?

Amazon API Gateway, Amazon S3, Amazon CloudFront, AWS CodeBuild, AWS CodePipeline, Amazon CloudWatch, OpenSearch, Amazon Bedrock, Amazon SageMaker Jumpstart, Hugging Face.

What results were reported?

Monthly internal users: over 1,000; Average interaction time per user: 5 minutes per user; Platform build time: under 6 weeks; Time to reach 1,000 monthly users: under a month (source-reported, not independently verified).

How is this back office ops AI workflow structured?

Employee accesses tile interface → RAG retrieves relevant chunks → LLM generates response → Answers delivered to employees → New use cases surface and feed back.

Related back office ops cases