Cognite and ABB integrate agentic AI into industrial workflows

Cognite announced a collaboration with ABB to integrate an agentic layer into ABB's industrial applications, including ABB Ability™ SafetyInsight™ and ABB Ability™ AlarmInsight™, using the Cognite Industrial AI and Data platform, according to a Business Wire release published May 21, 2026. The release says the integration is intended to enable "agent-to-agent" orchestration so applications can autonomously interpret data, reason through logic, and trigger cross-system actions. The release identifies Aker BP as the first customer; Paula Doyle, Chief Digital Officer at Aker BP, is quoted saying the company aims to "deploy hundreds of agents by 2026." The release also cites Aker BP's current production efficiency at 96% and a production growth target of 525,000 barrels of oil equivalent per day by 2028. Gino Hernandez, Head of Global Digital Business at ABB's Energy Industries division, is quoted on ABB's commitment to supporting customers' move toward autonomous operations.

The Business Wire release characterizes the new capability as an "agentic orchestration" layer that connects disparate data sources and runs cross-system workflows such as multi-system risk assessments and alarm rationalization. The release claims these processes can be completed "an order of magnitude faster than manual coordination" when agent-to-agent interactions are used to reason and trigger actions across systems. The announcement frames the offering as combining ABB's domain-focused operational applications with Cognite's industrial data and AI tooling to break down data silos and enable outcome-based software orchestration.

Editorial analysis: Industrial operators and vendors have increasingly moved from analytics and dashboards to automation and closed-loop decision automation over the last three years. Companies and system integrators adopting agentic patterns are pursuing tighter runtime coordination between monitoring, control, and decision systems, while also emphasizing model explainability and safety in critical-path environments. Observers following industrial AI deployments note that alarm rationalization, cross-system incident response, and multi-system risk assessment are common initial use cases because they concentrate value and reduce operator cognitive load.

For practitioners: indicators to follow include:

• •Adoption metrics and scale, for example the number of production agents deployed and the degree of cross-system orchestration realized in live operations;
• •How vendors surface provenance, explainability, and safety controls for agent actions inside operational technology (OT) environments;
• •Performance claims versus measured outcomes, such as time-to-resolution improvements for alarm storms and measured reductions in operator interventions.

Editorial analysis: The next practical questions for engineers and integrators will be integration patterns for OT/IT boundaries, the tolerance for autonomous agent actions in safety-critical workflows, and the telemetry required to audit agent decisions. Public deployments and third-party measurements will be the clearest signals of whether the agentic approach yields reproducible operational improvements at scale.