Thomson Reuters Open Arena: No-Code AI for Professionals

Thomson Reuters (TR) has successfully democratized AI creation across its organization with Open Arena, a highly scalable, no-code AI solution powered by Amazon Bedrock and other AWS services. Recognizing AI’s transformative potential, TR developed Open Arena to enable non-technical professionals to build customized AI solutions, achieving 70% employee adoption and 19,000 monthly active users.

Open Arena supports thousands of self-served, customized AI solutions for internal workflows or TR products, embodying TR’s commitment to making cutting-edge AI accessible without coding expertise. Initially a rapid prototype, Open Arena evolved into an enterprise solution built on the TR AI Platform, offering secure, scalable, and standardized services covering the entire AI development lifecycle, significantly accelerating time to production. It addresses critical enterprise AI challenges: enabling consistent LLM experiences for diverse users, ensuring security and quality through evaluation and governance, boosting speed and reusability with automated workflows, and managing resources and costs transparently.

The platform caters to various user types—from “AI explorers” summarizing documents and engaging in LLM chats, to “AI creators” and Subject Matter Experts (SMEs) building complex workflows, and developers deploying new AI capabilities. Key features include an intuitive no-code interface with visual workflow builders and pre-built templates, seamless integration with TR’s Enterprise AI Platform (SSO-enabled, enhanced data security), and robust solution management with a searchable repository and usage analytics.

Technically, Open Arena leverages Amazon Bedrock for foundational model access (supporting models from AI21 Labs, Anthropic, Meta, Amazon, etc.) and utilizes Bedrock Flows as a core feature of its “Chain Builder” (v2). This allows users to drag-and-drop components like prompts, agents, and knowledge bases to create sophisticated AI workflows without coding. The architecture employs a cross-account design for stringent data isolation, with orchestration managed centrally and user-specific data residing in dedicated AWS accounts, ensuring compliance and security. Other AWS services like OpenSearch, S3, DynamoDB, Lambda, API Gateway, and Step Functions underpin its serverless, scalable infrastructure, dynamically allocating resources to optimize performance and cost.

The evolution from Open Arena v1 (Classic) to v2 (Chain Builder) significantly expanded capabilities, offering advanced visual workflow design and parallel execution. TR plans to further enhance Open Arena with custom components, new templates, collaboration features, multimodal capabilities, and expanded use cases, continuing its mission to empower professionals with generative AI.

(Source: https://aws.amazon.com/blogs/machine-learning/democratizing-ai-how-thomson-reuters-open-arena-supports-no-code-ai-for-every-professional-with-amazon-bedrock/)