Company CrossnoKaye deploys AI agent AURA on Amazon Bedrock AgentCore to control industrial operations
The company CrossnoKaye runs the AI agent AURA on the Amazon Bedrock AgentCore platform to control more than 100 industrial facilities and 13 000 assets. According to the company, this increases production by 11–17% and reduces energy costs by an average of 21%.
The company CrossnoKaye operates ATLAS, a software-defined automation platform that combines live data from operations with the AI reasoning agent AURA and closed-loop equipment control in food and beverage operations, industrial refrigeration, quarries, and manufacturing plants. The platform covers more than 100 facilities and 13 000 controlled assets, stores more than 22 trillion data points, and ingests more than 50 billion new measurements daily; it runs more than 1 000 applications for advanced process control (APC). According to the company, these applications increase production output by 11–17% and reduce overall facility energy costs by an average of 21%.
The architecture is divided into three layers. At the facility level, local Kubernetes clusters with redundant hardware run equipment control (boilers, crushers, furnaces, refrigeration and compressor equipment) directly through existing PLC and DCS infrastructure without the need for hardware replacement; communication with the cloud is outbound-only via Amazon VPC Lattice and Amazon MSK, with no open inbound ports into the facility network. In the cloud, a continuous optimization pipeline combining physical models, model predictive control, reinforcement learning, energy tariffs, and weather forecasts runs, computing optimal settings and dispatching them via Amazon SQS FIFO queues — according to the company, this amounts to 650 000 closed-loop control actions daily. The AURA agent itself runs on Amazon EKS with Temporal workflows, and for numerical computations on sensor data it uses the code interpreter tool within Amazon Bedrock AgentCore, which executes Python code in isolated AWS microVM sandboxes; Amazon Bedrock Guardrails and local safety checks on edge devices are used in addition.
The company further states more than 9 000 control software deployments per year and more than 1 million operating hours without a security incident. The source article ends at this point; details of how AURA specifically uses interactive chats, repeatable tasks, and automated routines are not available. See the source article for details.
Why it matters
The case demonstrates a concrete architecture for integrating an LLM agent into the control of physical industrial processes without the need to change existing PLC/DCS/SCADA infrastructure and without scattering credentials across external tools — numerical computations are separated into isolated sandboxes, and local control continues even if the cloud connection fails. For companies in heavy industry, this serves as a reference point when evaluating offerings from vendors of AI agents for OT environments, including questions of safety and reliability during operation.
Relevant practical impact
What this means
For a business
The company demonstrates a deployment model of an AI agent directly into closed-loop operational control, with isolation of risky computations and a local safety fallback; the declared benefits (11–17% increase in production, average 21% energy savings) are vendor claims, not an independently verified result.
ProcessesCheck the original
Event sources
only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.