AI Teammates: how monday.com runs production AI agents on Amazon Bedrock

Monday.com has developed a sophisticated system of AI teammates, or agentic AI, that operate within its existing infrastructure, powered by Amazon Bedrock. This system enables real-time, end-to-end automation of software development tasks, with agents acting as co-workers integrated into daily workflows. The architecture involves a combination of AWS services, including SNS, SQS, EKS, RDS, ElastiCache, EFS, and S3, to manage event flow, state, and model calls efficiently.

The company has implemented a multi-layered approach to AI adoption, progressing from AI-assisted coding to fully autonomous agents capable of owning entire delivery processes. These agents are embedded into monday.com's microservices environment, with a focus on safety, auditability, and seamless integration with human teams.

Significant technical innovations include the use of Amazon Bedrock for model hosting, a custom runtime wrapper for model calls, and a robust system of evaluation and guardrails to ensure quality and safety. The deployment strategy emphasizes reliability, scalability, and compliance, with agents operating within a shared state and identity framework aligned with monday.com's existing infrastructure.

This implementation demonstrates how large-scale enterprise software can incorporate advanced AI agents without disrupting existing workflows, providing a blueprint for similar organizations aiming to scale AI-driven automation. The approach also highlights the importance of operational feedback loops, automated review systems, and confidence scoring to enable safe, autonomous code merging.

The impact of this development extends to the broader AI community by illustrating practical, scalable solutions for deploying agentic AI in production environments. It underscores the potential for AI to transform software engineering and enterprise operations, emphasizing the importance of infrastructure, safety, and integration in achieving this goal.