Kyndryl launches agentic AI service management toolkit

Kyndryl has launched Agentic Service Management, an offering aimed at enterprises adopting autonomous AI in IT operations.

The product combines a maturity model, structured assessments and implementation blueprints to help companies move from traditional service operations to more automated workflows. It is designed to provide a control layer for agentic AI by defining how autonomous systems are authorised, constrained, monitored and corrected in real time.

The launch targets a problem becoming more visible as AI agents move from pilot projects into production. In sectors such as healthcare, financial services and government, organisations are starting to use software agents that can make decisions and execute tasks with limited human intervention. Many governance models, however, still reflect slower, more predictable software environments.

Kyndryl argues this creates an accountability gap between what autonomous systems can do and what enterprise operating models can safely supervise. Many businesses, it says, are trying to deploy agentic AI without clear authority structures, escalation paths or operational oversight suited to systems that can act independently.

The service is being offered through Kyndryl Consult. The assessment reviews an organisation's current position across service management, AI governance, security and operations. It compares existing policies, controls and workflows against relevant standards and frameworks, including ISO 42001, to assess readiness for agentic operations.

Customers then receive a gap analysis and a phased adoption roadmap. The model is intended to support the use of autonomous systems in cloud-native and AI-native environments while maintaining human oversight for governance, risk and service outcomes.

Kyndryl has also positioned the launch alongside a separate service, Agentic AI Digital Trust. That offering is described as a framework for governing AI agents, reducing risk and managing deployments across hybrid and multi-cloud environments, with a focus on regulated sectors where data protection, compliance and classification are central concerns.

Governance gap

The launch comes amid a broader debate across the technology industry about how to control AI systems that do more than generate content or analyse data. Agentic AI refers to software that can pursue goals, trigger actions and interact with systems with some degree of autonomy. That can include handling IT incidents, reconfiguring infrastructure or executing workflow decisions across connected platforms.

For large organisations, the attraction is speed and scale. The challenge is accountability when software agents act across complex estates spanning on-premise systems, public cloud services and sector-specific regulatory requirements.

Kyndryl cited findings from its Readiness Report showing that more than two-thirds of organisations are investing heavily in AI, while nearly half struggle to achieve meaningful returns. It linked that gap to governance, workflows and controls that remain rooted in earlier operating models.

"Most enterprise environments were built for people running tickets and tools, not for fleets of autonomous agents executing tasks across hybrid and multi-cloud estates-and this mismatch is limiting AI from moving out of pilots to outcomes," said Kris Lovejoy, Global Head of Strategy at Kyndryl. "You can't scale agentic workflows on top of operating models that were designed for manual work. Organisations need clear controls, repeatable practices and measurable stages of adoption so AI agents can act autonomously where appropriate-while people remain accountable for governance, risk and service outcomes," added Lovejoy.

Internal use

Kyndryl is also applying the approach within its own IT service delivery operations. Through its Kyndryl Bridge platform, some related functions are already available to customers, where they are used to add operational intelligence and support staff responsible for oversight and decision-making across critical systems.

Its broader automation base currently executes nearly 200 million automations each month through more than 8,000 certified playbooks. That installed base offers some indication of the operational environment in which Kyndryl is seeking to embed more autonomous software behaviour while keeping existing service management disciplines in place.

Market shift

The launch reflects a broader shift in enterprise technology services as suppliers move from advising clients on AI strategy to designing practical governance models for live deployment. Many businesses have spent the past two years testing generative AI tools, but the move to software agents that can take action inside production systems raises a different set of operational questions.

These include who approves agent behaviour, how exceptions are handled, what controls apply when an agent interacts with sensitive systems, and how responsibility is assigned when something goes wrong. Service management, long associated with ticketing, change control and incident processes, is now being reframed as one possible mechanism for supervising these systems.

Kyndryl's position is that agentic AI needs a formal operating model before it can be widely used in enterprise IT.