- Understand agents, serving as a single source of truth to help mitigate the risks associated with shadow AI.
- Validate each agent before launch by testing for security, resilience, and policy compliance to ensure they meet your standards before going live.
- Maintain control with real-time guardrails that keep agents operating within approved boundaries.
Security testing, validation, and threat modeling should be incorporated into development pipelines, Kyndryl stated. “Additionally, runtime protections such as anomaly detection, guardian agents, and rapid isolation capabilities can help contain incidents before they escalate. By making security and governance foundational rather than treating them as afterthoughts, organizations can confidently scale agentic AI, knowing that risks are proactively managed, and trust is maintained with customers, partners, and regulators,” Kyndryl stated.
The new service is just one of the platforms the vendor offers to manage AI agents. Last year Kyndral introduced its Agentic AI Framework. That package offers an orchestration system built to deploy and manage autonomous, self-learning agents across business workflows in on-prem, cloud, or hybrid IT environments, according to the company.
Specialized agents are deployed to gather IT information, such as data analysis, compliance checks, incident response or service desk ticket resolution. Over time, agents learn from data and outcomes to improve decision-making and adapt workflows autonomously, and an orchestration engine parses that data to let enterprise systems adjust to changing conditions in real time, Kyndryl stated. The platform defines what actions agents can and cannot do, basically setting policy across the enterprise.
